| *** naohirot has joined #openstack-ironic | 00:02 | |
| *** marcoemorais1 has joined #openstack-ironic | 00:02 | |
| *** marcoemorais1 has quit IRC | 00:02 | |
| *** marcoemorais1 has joined #openstack-ironic | 00:03 | |
| *** marcoemorais has quit IRC | 00:05 | |
| *** igordcard has joined #openstack-ironic | 00:09 | |
| *** zyluo_ has joined #openstack-ironic | 00:14 | |
| *** marcoemorais1 has quit IRC | 00:15 | |
| *** marcoemorais has joined #openstack-ironic | 00:15 | |
| *** igordcard has quit IRC | 00:17 | |
| *** penick has quit IRC | 00:18 | |
| *** achanda has joined #openstack-ironic | 00:20 | |
| *** achanda has quit IRC | 00:34 | |
| *** marcoemorais has quit IRC | 00:35 | |
| *** marcoemorais has joined #openstack-ironic | 00:36 | |
| *** marcoemorais has quit IRC | 00:36 | |
| *** marcoemorais has joined #openstack-ironic | 00:36 | |
| *** marcoemorais has quit IRC | 00:37 | |
| *** marcoemorais has joined #openstack-ironic | 00:37 | |
| *** zyluo_ has quit IRC | 00:39 | |
| *** achanda has joined #openstack-ironic | 00:47 | |
| yjiang5 | JayF: hi, what's the username/passwd for the IPA deploy kernel/ramdisk provided in the devstack? | 00:49 |
|---|---|---|
| *** marcoemorais has quit IRC | 00:50 | |
| *** marcoemorais has joined #openstack-ironic | 00:50 | |
| *** anderbubble has joined #openstack-ironic | 00:57 | |
| *** anderbubble has quit IRC | 00:59 | |
| *** ryanpetrello has joined #openstack-ironic | 00:59 | |
| *** achanda has quit IRC | 01:08 | |
| *** nosnos has joined #openstack-ironic | 01:09 | |
| *** achanda has joined #openstack-ironic | 01:18 | |
| *** killer_prince has joined #openstack-ironic | 01:29 | |
| *** killer_prince is now known as lazy_prince | 01:29 | |
| *** chenglch has joined #openstack-ironic | 01:40 | |
| *** achanda has quit IRC | 01:41 | |
| *** achanda has joined #openstack-ironic | 01:42 | |
| *** achanda has quit IRC | 01:46 | |
| *** chenglch has quit IRC | 01:49 | |
| *** ryanpetrello has quit IRC | 01:50 | |
| *** victor_lowther__ has quit IRC | 01:56 | |
| *** victor_lowther__ has joined #openstack-ironic | 01:56 | |
| *** vdrok has quit IRC | 01:57 | |
| *** vdrok has joined #openstack-ironic | 01:57 | |
| *** dlaube has quit IRC | 02:01 | |
| openstackgerrit | Zhongyue Luo proposed a change to openstack/ironic-specs: Dnsmasq DHCP service provider https://review.openstack.org/132744 | 02:08 |
| *** mitz_ has joined #openstack-ironic | 02:14 | |
| JayF | yjiang5: if you append "coreos.autologin" to the kernel command line, it should automatically login. Alternatively, bake an ssh key into the image. | 02:20 |
| *** Viswanath has joined #openstack-ironic | 02:23 | |
| *** marcoemorais has quit IRC | 02:26 | |
| *** Viswanath has quit IRC | 02:26 | |
| *** cppforlife has joined #openstack-ironic | 02:42 | |
| *** subscope has quit IRC | 02:58 | |
| *** spandhe has quit IRC | 03:16 | |
| *** nosnos has quit IRC | 03:20 | |
| *** penick has joined #openstack-ironic | 03:53 | |
| naohirot | Hello Ironic team, I recently joined the community. | 04:03 |
| naohirot | It took for a couple of weeks to realize that past logs are available at http://eavesdrop.openstack.org/irclogs/%23openstack-ironic. | 04:03 |
| naohirot | Now I'm reading them :-). | 04:04 |
| *** shakamunyi has joined #openstack-ironic | 04:04 | |
| *** achanda has joined #openstack-ironic | 04:11 | |
| *** nosnos has joined #openstack-ironic | 04:12 | |
| *** penick has quit IRC | 04:13 | |
| *** subscope has joined #openstack-ironic | 04:24 | |
| *** pensu has joined #openstack-ironic | 04:25 | |
| *** achanda_ has joined #openstack-ironic | 04:34 | |
| *** achanda has quit IRC | 04:36 | |
| *** shakamun_ has joined #openstack-ironic | 04:45 | |
| *** shakamunyi has quit IRC | 04:49 | |
| *** rushiagr_away is now known as rushiagr | 04:51 | |
| *** pensu has quit IRC | 04:56 | |
| *** pensu has joined #openstack-ironic | 04:59 | |
| *** pcrews has quit IRC | 05:10 | |
| *** pensu has quit IRC | 05:15 | |
| *** chenglch has joined #openstack-ironic | 05:17 | |
| *** spandhe has joined #openstack-ironic | 05:25 | |
| *** spandhe_ has joined #openstack-ironic | 05:28 | |
| *** spandhe has quit IRC | 05:30 | |
| *** spandhe_ is now known as spandhe | 05:30 | |
| *** chenglch has quit IRC | 05:32 | |
| *** marcoemorais has joined #openstack-ironic | 05:36 | |
| *** marcoemorais1 has joined #openstack-ironic | 05:38 | |
| *** rakesh_hs has joined #openstack-ironic | 05:38 | |
| *** k4n0 has joined #openstack-ironic | 05:39 | |
| *** subscope has quit IRC | 05:40 | |
| *** marcoemorais has quit IRC | 05:41 | |
| *** chenglch|2 has joined #openstack-ironic | 05:44 | |
| *** pensu has joined #openstack-ironic | 05:48 | |
| *** foexle has joined #openstack-ironic | 06:26 | |
| *** achanda_ has quit IRC | 06:27 | |
| *** achanda has joined #openstack-ironic | 06:29 | |
| openstackgerrit | Harshada Mangesh Kakad proposed a change to openstack/ironic: Add serial console feature to seamicro driver https://review.openstack.org/132628 | 06:36 |
| *** dlpartain has joined #openstack-ironic | 06:44 | |
| *** spandhe has quit IRC | 06:51 | |
| *** datajerk has quit IRC | 06:58 | |
| *** rushiagr is now known as rushiagr_away | 07:11 | |
| *** dlpartain has quit IRC | 07:20 | |
| *** dlpartain has joined #openstack-ironic | 07:20 | |
| *** achanda has quit IRC | 07:20 | |
| *** rameshg87 has joined #openstack-ironic | 07:22 | |
| *** yuanying has quit IRC | 07:32 | |
| *** adam_g` has joined #openstack-ironic | 07:33 | |
| *** jerryz has joined #openstack-ironic | 07:35 | |
| *** adam_g has quit IRC | 07:38 | |
| *** yuanying has joined #openstack-ironic | 07:44 | |
| *** rushiagr_away has quit IRC | 07:49 | |
| *** yuanying has quit IRC | 07:49 | |
| *** foexle has quit IRC | 07:53 | |
| *** jerryz has quit IRC | 07:56 | |
| *** yuanying has joined #openstack-ironic | 07:58 | |
| *** yuanying has quit IRC | 08:05 | |
| *** yuanying has joined #openstack-ironic | 08:07 | |
| *** yuanying has quit IRC | 08:09 | |
| *** yuanying has joined #openstack-ironic | 08:14 | |
| *** foexle has joined #openstack-ironic | 08:15 | |
| *** dlpartain has quit IRC | 08:15 | |
| *** yuanying has quit IRC | 08:19 | |
| *** achanda has joined #openstack-ironic | 08:21 | |
| *** ndipanov_gone is now known as ndipanov | 08:21 | |
| GheRivero | morning @ironic | 08:22 |
| *** rushiagr has joined #openstack-ironic | 08:25 | |
| *** achanda has quit IRC | 08:26 | |
| *** Nisha has joined #openstack-ironic | 08:28 | |
| *** yuanying has joined #openstack-ironic | 08:34 | |
| *** yuanying has quit IRC | 08:38 | |
| *** ifarkas has joined #openstack-ironic | 08:44 | |
| *** yuanying has joined #openstack-ironic | 08:44 | |
| *** yuanying has quit IRC | 08:50 | |
| *** yuanying has joined #openstack-ironic | 08:50 | |
| *** chenglch|2 has quit IRC | 08:53 | |
| *** chenglch|2 has joined #openstack-ironic | 08:54 | |
| *** jistr has joined #openstack-ironic | 08:59 | |
| *** yuanying has quit IRC | 09:00 | |
| *** derekh has joined #openstack-ironic | 09:01 | |
| *** rameshg87 has quit IRC | 09:03 | |
| *** Marga_ has joined #openstack-ironic | 09:05 | |
| *** Marga_ has quit IRC | 09:05 | |
| *** Marga_ has joined #openstack-ironic | 09:05 | |
| *** yuanying has joined #openstack-ironic | 09:06 | |
| *** yuanying has quit IRC | 09:08 | |
| *** chenglch|2 has quit IRC | 09:08 | |
| *** jiangfei has joined #openstack-ironic | 09:10 | |
| *** dlpartain has joined #openstack-ironic | 09:13 | |
| *** yuanying has joined #openstack-ironic | 09:15 | |
| *** yuanying has quit IRC | 09:17 | |
| *** rameshg87 has joined #openstack-ironic | 09:18 | |
| *** lucasagomes has joined #openstack-ironic | 09:19 | |
| *** zhangyang_ has joined #openstack-ironic | 09:22 | |
| *** lazy_prince has quit IRC | 09:22 | |
| Nisha | Hi lucasagomes | 09:30 |
| *** yuanying has joined #openstack-ironic | 09:31 | |
| *** athomas has joined #openstack-ironic | 09:33 | |
| *** yuanying has quit IRC | 09:33 | |
| *** yuanying has joined #openstack-ironic | 09:36 | |
| *** MattMan has quit IRC | 09:37 | |
| *** yuanying has quit IRC | 09:37 | |
| *** MattMan has joined #openstack-ironic | 09:38 | |
| *** Marga_ has quit IRC | 09:43 | |
| *** yuanying has joined #openstack-ironic | 09:43 | |
| lucasagomes | Nisha, good morning/afternoon | 09:45 |
| *** igordcard has joined #openstack-ironic | 09:46 | |
| *** mkerrin has quit IRC | 09:47 | |
| Nisha | lucasagomes: hi | 09:54 |
| Nisha | lucasagomes: saw Dmitry's suggestion for properties discovery as discussed in summit. | 09:55 |
| Nisha | lucasagomes: however i had some queries | 09:55 |
| Nisha | lucasagomes: 1. When we say discover shall be a new state, then it shall be implemented as we have power/deploy today? | 09:56 |
| Nisha | lucasagomes: secondly, is it necessary to do properties discovery async always? means a driver may want to do synchronously | 09:57 |
| openstackgerrit | Harshada Mangesh Kakad proposed a change to openstack/ironic: Add serial console feature to seamicro driver https://review.openstack.org/132628 | 09:58 |
| lucasagomes | Nisha, yeah there's still some talking need because the session we had (IMO) went a bit out of hands | 09:58 |
| lucasagomes | for 1. yup, lemme show u the draft of the state machine | 09:58 |
| *** leopoldj has joined #openstack-ironic | 09:58 | |
| Nisha | lucasagomes: ok | 09:58 |
| lucasagomes | 2. (Assuming ur talking about out of band discovery) I believe that people will push hard to us to have it async, there's a strong opnion in the Ironic community that everything that touches the BMC should be done async | 09:59 |
| Nisha | yes, | 09:59 |
| Nisha | lucasagomes: hmmm but it can be left to driver/user also | 10:00 |
| Nisha | lucasagomes: either way, the conductor will be able to honor it\ | 10:00 |
| *** romcheg has joined #openstack-ironic | 10:01 | |
| Nisha | lucasagomes: i am talking about generic node-properties spec | 10:01 |
| lucasagomes | Nisha, https://plus.google.com/photos/+LucasAlvaresGomes/albums/6079343647897766897/6079343648628836562?pid=6079343648628836562&oid=111472618817384816105 | 10:01 |
| Nisha | lucasagomes: this is from summit ? :) | 10:02 |
| *** pelix has joined #openstack-ironic | 10:02 | |
| lucasagomes | Nisha, yup | 10:02 |
| Nisha | lucasagomes: :) | 10:03 |
| lucasagomes | Nisha, right yeah, one good thing about the generic one is that we don't need to care about in-band anymore | 10:03 |
| lucasagomes | in-band will be a separated service | 10:03 |
| *** sambetts has joined #openstack-ironic | 10:03 | |
| lucasagomes | so we can focus on making it generic, cross-vendor, for the out-of-band | 10:03 |
| Nisha | ok | 10:03 |
| lucasagomes | Nisha, about left the driver choose in-band or out-of-band hmm it's possible, but wouldn't be nicer to be kinda standard one or another | 10:04 |
| lucasagomes | I see sync being easier, but for DRAC at least (AFAIK) we need to issue a couple of requests to the BMC | 10:04 |
| lucasagomes | so it has to be async | 10:04 |
| lucasagomes | it's not only 1 :( | 10:04 |
| Nisha | lucasagomes: driver choosing in-band? i didnt get.... do u mean sync and async by this? | 10:05 |
| lucasagomes | Nisha, no in-band is out of ironic codebase | 10:05 |
| lucasagomes | if people is willing to do in-band discovery they will need a third-party service to | 10:05 |
| Nisha | yes , i was saying about this statement "<lucasagomes> Nisha, about left the driver choose in-band or out-of-band hmm it's possible" | 10:05 |
| *** romcheg has quit IRC | 10:06 | |
| lucasagomes | ohhh | 10:06 |
| lucasagomes | sync and async | 10:06 |
| lucasagomes | my bad | 10:06 |
| Nisha | :) | 10:06 |
| lucasagomes | haha early morning | 10:06 |
| Nisha | :) | 10:06 |
| lucasagomes | mixing words in my head here | 10:06 |
| Nisha | so i was just thinking to propose discovery rpcapi to be sync and async similar to what u have proposed for vendor_passthru | 10:07 |
| Nisha | ;) | 10:07 |
| Nisha | copy-paste ;) | 10:07 |
| Nisha | does it sounds ok? | 10:07 |
| *** jiangfei has quit IRC | 10:08 | |
| lucasagomes | Nisha, I think so, but as an alert I think that it's going to get some push-backs, for iLO it may not be a problem but I believe that the overall understanding of BMCs is that it kinda sucks and can get jammed and the request would timed out | 10:09 |
| lucasagomes | so things like that should be async | 10:09 |
| *** dlpartain has left #openstack-ironic | 10:10 | |
| lucasagomes | but no problems in proposing it | 10:10 |
| lucasagomes | I may be wrong | 10:10 |
| *** ndipanov has quit IRC | 10:10 | |
| *** romcheg has joined #openstack-ironic | 10:12 | |
| *** ndipanov has joined #openstack-ironic | 10:12 | |
| *** yuanying has quit IRC | 10:14 | |
| *** jiangfei has joined #openstack-ironic | 10:19 | |
| *** yuanying has joined #openstack-ironic | 10:20 | |
| sambetts | Morning lucasagomes | 10:21 |
| lucasagomes | sambetts, yo morning | 10:21 |
| *** achanda has joined #openstack-ironic | 10:23 | |
| naohirot | Good morning lucasagomes, | 10:27 |
| lucasagomes | naohirot, good morning there | 10:28 |
| naohirot | I have question regarding DRAC | 10:28 |
| naohirot | hi | 10:28 |
| *** achanda has quit IRC | 10:28 | |
| *** jiangfei_ has joined #openstack-ironic | 10:28 | |
| naohirot | do you have a document such us the iLO driver? | 10:28 |
| naohirot | I'll paste the URL | 10:29 |
| naohirot | https://wiki.openstack.org/wiki/Ironic/Drivers/iLODrivers | 10:29 |
| lucasagomes | no :( I have to come up with some documentation for it. The only work I did on the DRAC was the management interface | 10:29 |
| Nisha | lucasagomes: thanks, i will propose it and then lets see if it goes through | 10:29 |
| lucasagomes | we will work more on the driver this cycle to support shiny new features like iLO does now | 10:30 |
| Nisha | lucasagomes: thanks :) | 10:30 |
| lucasagomes | Nisha, awesome! thanks | 10:30 |
| naohirot | I see. | 10:30 |
| naohirot | so I have very basic question. | 10:30 |
| Nisha | lucasagomes: i said thanks for following : "we will work more on the driver this cycle to support shiny new features like iLO does now" | 10:30 |
| Nisha | :) | 10:31 |
| lucasagomes | :) | 10:31 |
| naohirot | If user use pxe_drac driver | 10:31 |
| *** dtantsur|afk is now known as dtantsur | 10:31 | |
| *** jiangfei has quit IRC | 10:31 | |
| dtantsur | Morning Ironic | 10:31 |
| sambetts | Morning dtantsur | 10:31 |
| * dtantsur is trying to wake up... | 10:31 | |
| Nisha | dtantsur: : morning | 10:31 |
| naohirot | what kind of benefit does user have? | 10:32 |
| dtantsur | sambetts, Nisha, hi! | 10:32 |
| Nisha | dtantsur: i left a ques for you in discovery spec ... | 10:32 |
| dtantsur | lemme see | 10:32 |
| Nisha | i will catch it up in my night time | 10:32 |
| Nisha | i have to leave office now | 10:32 |
| naohirot | iLO driver seems to provides some security benefit. | 10:33 |
| lucasagomes | naohirot, right now, I don't think you have any benefits of using drac_pxe or ipmitool_pxe... But that said, we have specs up for DRAC to do auto discovery of the properties and RAID configuration | 10:33 |
| Nisha | just let me know your comments, i will then update spec and post it | 10:33 |
| dtantsur | Nisha, ack | 10:33 |
| lucasagomes | naohirot, we want to add support for virtual media deployment too | 10:33 |
| lucasagomes | naohirot, indeed it does | 10:33 |
| lucasagomes | naohirot, the HP guys are doing great supporting their hardware, the main person that was working on the DRAC from red hat last cycle had to step out a bit | 10:34 |
| naohirot | I see, so far auto discovery and RAID configuration are the target? | 10:34 |
| lucasagomes | but he's back this cycle so we will try to catch up | 10:34 |
| lucasagomes | naohirot, yup, and I believe virtual media | 10:34 |
| naohirot | Okay | 10:35 |
| naohirot | Another basic question | 10:35 |
| naohirot | Why doesn't Ironic mention anything about SMASH CLP? | 10:35 |
| naohirot | IPMI is enough for standard? | 10:36 |
| naohirot | standard -> standard I/F. | 10:36 |
| *** Nisha has quit IRC | 10:37 | |
| lucasagomes | well nobody proposed using it yet... and I believe most of the use cases for Ironic right now are happy with IPMI | 10:37 |
| lucasagomes | I mean not happy but that's the only thing they have | 10:37 |
| lucasagomes | the biggest deployment of ironic AFAIK uses open compute so they only have vanilla IPMI to manage their machines | 10:38 |
| naohirot | Okay | 10:38 |
| lucasagomes | and we just graduated, the last cycle the focus were most targeting graduation | 10:38 |
| lucasagomes | so we did ton of works on the nova driver (that was the biggest hurdles) and tests (granade, tempest, migration) | 10:38 |
| *** yuanying has quit IRC | 10:39 | |
| *** yuanying has joined #openstack-ironic | 10:39 | |
| naohirot | Yes, I understand. Ironic is still young. I have further basic question. | 10:39 |
| naohirot | I still didn't understand the difference between pxe_ipmitool and agent_ipmitool. | 10:40 |
| lucasagomes | yeah the names are a bit misleading there | 10:41 |
| lucasagomes | but the difference is the way it's deployed | 10:41 |
| naohirot | I believe that both are PXE boots. | 10:41 |
| lucasagomes | naohirot, yes, we def need to naming things better... lemem try to explain | 10:41 |
| lucasagomes | so for the "pxe_" namespace it mostly means PXE + ISCSI | 10:42 |
| naohirot | Agent install physical OS into local disk, on the other hand PXE didn't install anything into local disk. | 10:42 |
| naohirot | Am I correct? | 10:42 |
| lucasagomes | where a ramdisk will be booted, it will then expose the disk via ISCSI and the ironic conductor will copy the image (via dd) onto that disk | 10:42 |
| *** marcoemorais1 has quit IRC | 10:42 | |
| lucasagomes | the "agent_" uses a more smart ramdisk that will boot on the machine and will fetch the image directly from glance/swift and copy it to the disk there | 10:43 |
| lucasagomes | naohirot, both installs things on the local disk | 10:43 |
| lucasagomes | the incosistency here is that "pxe_" always assume that the image is a partition image | 10:43 |
| lucasagomes | so even after deployed it will PXE boot the kernel+ramdisk to boot the machine | 10:44 |
| naohirot | Okay, both installs physical OS into local disk. | 10:44 |
| lucasagomes | where "agent_" assumed it's a full disk image, with a bootloader | 10:44 |
| lucasagomes | yes | 10:44 |
| lucasagomes | we had a session in Paris about making it consistent | 10:44 |
| lucasagomes | this is a bug right now, because both drivers behave difently and it's not exposed by the user | 10:45 |
| lucasagomes | naohirot, https://etherpad.openstack.org/p/kilo-ironic-exposing-different-capabilities | 10:45 |
| lucasagomes | this is the etherpad from the summit | 10:45 |
| naohirot | Okay, in case of pxe_, user have to make partition by hands. | 10:45 |
| lucasagomes | naohirot, not really, the Ironic conductor will partition the disk | 10:45 |
| lucasagomes | over ISCSI | 10:46 |
| naohirot | but in case of agent_, agent takes care of partition. | 10:46 |
| *** yuanying has quit IRC | 10:46 | |
| naohirot | Okay, in case of pxe_, physical server must support iSCSI? | 10:46 |
| lucasagomes | both of them will take care of the partition, but the "pxe_" always do network boot because the image does not contain a bootloader (it's a partition image) nor the ironic conductor installs a boot loader | 10:47 |
| lucasagomes | naohirot, yeah, the ramdisk built will include all the iscsi modules/clients | 10:47 |
| naohirot | Aha, I see. | 10:47 |
| * lucasagomes realizes how much we need to make it consistent asap | 10:48 | |
| naohirot | Right now, physical server has to have local disks. If there is no local disk, what happens? | 10:49 |
| *** Haomeng has joined #openstack-ironic | 10:49 | |
| lucasagomes | naohirot, yup right now we assume they do | 10:50 |
| lucasagomes | naohirot, deploy fails :) | 10:50 |
| *** Haomeng|2 has quit IRC | 10:50 | |
| lucasagomes | naohirot, jroll was looking into having diskless deployments | 10:50 |
| lucasagomes | someone have raised it in the conference too | 10:50 |
| naohirot | I see, how about if physical server had HBA card and SAN. | 10:51 |
| naohirot | Is it recognized as local disk? | 10:51 |
| *** ndipanov has quit IRC | 10:52 | |
| naohirot | I see, it seems there are a lot of interesting discussion. I wanted to be there :-) | 10:54 |
| lucasagomes | yeah I haven't tried that... If the disk is exposed as a local disk perhaps that may work | 10:55 |
| lucasagomes | yeah it was pretty interesting indeed | 10:56 |
| naohirot | I see, I'll try later the case of SAN. | 10:57 |
| lucasagomes | there are many use cases, it's good that our drivers interface are flexible to support that (and we should work on some consistency mechanism) | 10:57 |
| lucasagomes | but we need people to code | 10:57 |
| lucasagomes | and review stuff | 10:57 |
| lucasagomes | naohirot, awesome | 10:57 |
| naohirot | yes, I'll report the result if we tried. | 10:58 |
| lucasagomes | ta much | 10:59 |
| naohirot | oh, another question. drac and ilo have separate python package of SOAP/XML part. | 10:59 |
| naohirot | Is it decision based on some policy? | 11:00 |
| naohirot | In case of iRMC, do I have to separate SOAP/XML as external library? | 11:01 |
| openstackgerrit | A change was merged to openstack/ironic: Remove redundant FunctionalTest usage in API tests https://review.openstack.org/133490 | 11:02 |
| lucasagomes | hmm I'm not sure cause I haven't started any of those drivers, DRAC uses python-wsman and iLO proliantutils | 11:02 |
| lucasagomes | rameshg87, ^ | 11:02 |
| *** yuanying has joined #openstack-ironic | 11:03 | |
| lucasagomes | if iRMC supports the same wsman or the other I don't see why not use it | 11:03 |
| lucasagomes | I mean one less cross-driver dependency | 11:03 |
| naohirot | It seems that iRMC http/https I/F is not pure SOAP. | 11:04 |
| naohirot | So wsman would not work, I believe. | 11:05 |
| *** yuriyz has quit IRC | 11:05 | |
| lucasagomes | right | 11:05 |
| *** ndipanov has joined #openstack-ironic | 11:05 | |
| lucasagomes | as a note, all those third-party libraries are not dependencies for Ironic itself | 11:05 |
| lucasagomes | operators should install those on their images depending on the driver they choose to use | 11:06 |
| naohirot | In case of iRMC, there is no way to get boot device and set boot device, and get list of boot devices. | 11:06 |
| naohirot | via SOAP. | 11:06 |
| naohirot | via http/https SOAP alike. | 11:07 |
| dtantsur | lucasagomes, hi, sorry to interrupt, but who and when is going to work on the state machine reform? do you have an idea? this thing kinda blocks everything :) | 11:07 |
| lucasagomes | naohirot, can it be done via ipmi? if so you may want to take a look at pyghmi as a lib for that | 11:07 |
| naohirot | Yes. | 11:08 |
| lucasagomes | dtantsur, I've no idea to be honest | 11:08 |
| lucasagomes | dtantsur, yup, it should be done soon, but I don't know if someone actually proposed to work on that | 11:08 |
| dtantsur | ah ok. I suppose it's devananda himself, but I maybe completely wrong :) | 11:08 |
| *** yuanying has quit IRC | 11:08 | |
| *** romcheg1 has joined #openstack-ironic | 11:08 | |
| lucasagomes | dtantsur, I think that it's open in the air for someone willing to do it | 11:08 |
| lucasagomes | I was assuming it would be kinda a plumbing for the decom work | 11:09 |
| *** romcheg has quit IRC | 11:09 | |
| dtantsur | well mostly, but it also blocks discovery | 11:09 |
| lucasagomes | indeed | 11:09 |
| dtantsur | lucasagomes, also, could you have a quick a look at https://review.openstack.org/#/c/100951/ to see if my comments there make any sense? | 11:10 |
| lucasagomes | +1 it's on my todo list | 11:10 |
| lucasagomes | just finishing something quicker here and I will take a look at it | 11:10 |
| *** yuanying has joined #openstack-ironic | 11:12 | |
| dtantsur | ack | 11:12 |
| naohirot | lucasagomes: In case of iRMC, I'm thinking to use SM CLP to get/set boot device form the management I/F. Do you think it's appropriate? | 11:12 |
| *** romcheg has joined #openstack-ironic | 11:13 | |
| naohirot | s/form/for | 11:13 |
| *** romcheg1 has quit IRC | 11:13 | |
| lucasagomes | naohirot, I don't see why not, specially if by using SM CLP enables you to have more advanced management control | 11:14 |
| *** yuanying has quit IRC | 11:15 | |
| naohirot | I see. The reason I asked is that SM CLP is a standard. so I just wandered. | 11:15 |
| lucasagomes | and also we will have some smash code :) maybe other drivers may be interested in using it in the future too | 11:15 |
| lucasagomes | sure | 11:16 |
| naohirot | I see. | 11:16 |
| naohirot | Thank you for sparing your time, lucasagomes! | 11:17 |
| naohirot | I'll read the references you kindly gave me. | 11:17 |
| lucasagomes | naohirot, no problem, ure very welcome | 11:17 |
| naohirot | Maybe I'll ask other questions tomorrow :-) | 11:18 |
| lucasagomes | +1 :D | 11:18 |
| naohirot | see you! | 11:18 |
| lucasagomes | see ya | 11:18 |
| *** sambetts_ has joined #openstack-ironic | 11:40 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Add sync and async support for passthru methods https://review.openstack.org/129662 | 11:42 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Make vendor methods discoverable via the Ironic API https://review.openstack.org/133533 | 11:42 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Vendor endpoints to support different HTTP methods https://review.openstack.org/129942 | 11:42 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Add a mechanism to route vendor methods https://review.openstack.org/129261 | 11:42 |
| *** sambetts has quit IRC | 11:42 | |
| rameshg87 | lucasagomes, just reading back | 11:46 |
| *** yuriyz has joined #openstack-ironic | 11:46 | |
| lucasagomes | rameshg87, morning/afternoon... no worries | 11:49 |
| rameshg87 | lucasagomes, morning :) | 11:50 |
| rameshg87 | lucasagomes, was away from my desk :) | 11:50 |
| lucasagomes | rameshg87, I see :D | 11:51 |
| *** naohirot has quit IRC | 11:51 | |
| rameshg87 | naohirot, we have proliantutils module that talks to the iLO using a protocol called RIBCL over HTTPS. basically sends in XML to iLO and gets back XML from iLO. we have made it as a python module: https://github.com/hpproliant/proliantutils/ | 11:52 |
| rameshg87 | oh i guess you quit right before i sent you the message | 11:52 |
| rameshg87 | lucasagomes, i am thinking we can do some testing on gate with virtual media with kvm/virtual box. | 11:53 |
| rameshg87 | lucasagomes, if we refactor virtual media out and make it implementation independent | 11:53 |
| rameshg87 | lucasagomes, basically it requires just a virtual media floppy and virtual media cdrom that kvm/virtual box already provides | 11:54 |
| rameshg87 | lucasagomes, what do you think ? | 11:54 |
| lucasagomes | rameshg87, you mean adding it to the SSH driver? | 11:54 |
| rameshg87 | lucasagomes, yes .. virtual media driver with ssh :) | 11:54 |
| lucasagomes | rameshg87, I think it's good, but I don't know exactly what are the benefits since we are not testing anything there | 11:55 |
| lucasagomes | I would love to have a IPMI listener service | 11:55 |
| rameshg87 | lucasagomes, is there one ? | 11:55 |
| lucasagomes | that would convert the IPMI commands to KVM/VBOX | 11:55 |
| *** andreykurilin has joined #openstack-ironic | 11:55 | |
| lucasagomes | because then at least we can test the contract with the IPMI commands see if they are correct etc | 11:55 |
| lucasagomes | rameshg87, not really :( not that I know at least | 11:56 |
| lucasagomes | but once we do that we can test the ipmi* drivers on gate too, and finally delete the ssh one | 11:56 |
| lucasagomes | cause it's just a work around really :/ | 11:56 |
| rameshg87 | lucasagomes, my thought we could test entire solution without actual hardware then - building the images, building the configs, uploading to swift, etc .. | 11:56 |
| lucasagomes | rameshg87, NobodyCam started working on something along tthat lines | 11:56 |
| rameshg87 | lucasagomes, all these don't play with real hardware | 11:56 |
| lucasagomes | rameshg87, oh, yeah indeed that would be important to test the generation of the ISO | 11:57 |
| rameshg87 | lucasagomes, yes | 11:57 |
| lucasagomes | yeah I see the benefit | 11:57 |
| lucasagomes | I wouldn't mind in having it in gate | 11:57 |
| rameshg87 | lucasagomes, yeah let me just think more about it and come up with a proposal | 11:57 |
| lucasagomes | +1 | 11:57 |
| rameshg87 | lucasagomes, are you working on boot-deploy interface separation ? | 11:58 |
| lucasagomes | rameshg87, not currently, I started looking into it and put some patches up | 11:58 |
| rameshg87 | lucasagomes, are they on gerrit ? | 11:58 |
| lucasagomes | but haven't finished the work | 11:58 |
| lucasagomes | rameshg87, inputs are welcome | 11:58 |
| lucasagomes | rameshg87, yup they are | 11:58 |
| rameshg87 | lucasagomes, would like to have a look :) | 11:58 |
| lucasagomes | rameshg87, https://review.openstack.org/#/c/127212/ | 11:59 |
| lucasagomes | starts there | 11:59 |
| rameshg87 | lucasagomes, got it | 11:59 |
| lucasagomes | but they are proof of concept | 11:59 |
| rameshg87 | lucasagomes, i have one question | 12:00 |
| rameshg87 | lucasagomes, do we need to make this mandatory that every driver should have boot and deploy separately ? | 12:00 |
| rameshg87 | lucasagomes, couldn't it be one way of making a deploy driver ? | 12:00 |
| rameshg87 | lucasagomes, what i meant was someone could still write a deploy driver with the before interfaces | 12:01 |
| lucasagomes | rameshg87, I believe that at the beggining we can't make it mandatory because drivers out of tree will break if we make it mandatory | 12:01 |
| rameshg87 | lucasagomes, yes | 12:01 |
| rameshg87 | lucasagomes, so boot interface is not mandatory for driver according to you ? | 12:01 |
| lucasagomes | but in the future, if we separate I think it makes sense to implement a boot interface and deploy interface separated | 12:02 |
| lucasagomes | rameshg87, well I haven't thought much about it to be honest | 12:02 |
| rameshg87 | lucasagomes, okay | 12:02 |
| lucasagomes | I understand that as a start point it can't be mandatory, but idk in the future it kinda makes sense to make people to divide the logic | 12:02 |
| lucasagomes | otherwise we lose the benefits of separating it | 12:03 |
| rameshg87 | lucasagomes, hmm ..okay .. i still wanted to discuss more on that .. | 12:03 |
| lucasagomes | +1 | 12:03 |
| rameshg87 | lucasagomes, but time to leave office :) | 12:03 |
| rameshg87 | lucasagomes, will join you later .. | 12:03 |
| lucasagomes | rameshg87, have a nice evening :) | 12:03 |
| rameshg87 | lucasagomes, have a good day :) | 12:04 |
| *** rameshg87 has quit IRC | 12:04 | |
| *** Marga_ has joined #openstack-ironic | 12:05 | |
| *** Guest9136 has joined #openstack-ironic | 12:09 | |
| *** Guest9136 is now known as annegentle | 12:10 | |
| *** nosnos has quit IRC | 12:13 | |
| *** nosnos has joined #openstack-ironic | 12:14 | |
| *** jiangfei_ has quit IRC | 12:15 | |
| *** nosnos has quit IRC | 12:18 | |
| *** jistr is now known as jistr|english | 12:23 | |
| soren | Is there a documented way to perform kernel updates for nodes managed by Ironic? | 12:24 |
| *** naohirot has joined #openstack-ironic | 12:29 | |
| *** pensu has quit IRC | 12:38 | |
| soren | Somehow I need to get the kernel and ramdisk refs updated in Nova's db... Then I need to get Ironic to reread that data and repopulate the tftpboot dir. | 12:44 |
| soren | Surely someone must have needed to update their kernels before? | 12:44 |
| *** andreykurilin has quit IRC | 12:56 | |
| *** andreykurilin has joined #openstack-ironic | 12:56 | |
| lucasagomes | soren, hi that's a good point. I don't think we have a mechanism for it | 12:58 |
| lucasagomes | I'm wondering whether rebuild would be the right way to do that | 12:58 |
| lucasagomes | maybe that's a bit overkill too | 12:58 |
| lucasagomes | soren, I don't understand the Nova's db part | 12:59 |
| soren | lucasagomes: Well, I tell nova which kernel/ramdisk I want to use. | 13:02 |
| soren | lucasagomes: ....and Nova passes that info on to Ironic somehow. | 13:02 |
| lucasagomes | soren, that's in the glance image right? | 13:02 |
| lucasagomes | in glance, the image has metadata about what is the kernel and ramdisk | 13:02 |
| soren | That's the typical approach, yes. | 13:02 |
| lucasagomes | in nova nova you say that you want to use that image, so Ironic looks at the glance image metadata and fetch the ramdisk/kernel | 13:02 |
| soren | You can specify them manually, but usually it's taken from the metadata of the fs image. | 13:02 |
| *** romcheg has quit IRC | 13:03 | |
| lucasagomes | right... | 13:03 |
| soren | Rebuild would destroy the local state, right? | 13:04 |
| lucasagomes | yeah it would since it reimage the node | 13:05 |
| lucasagomes | not ideal :/ | 13:05 |
| soren | Not useful, no. | 13:05 |
| lucasagomes | gotta think how it would work, you would upload a new kernel/ramdisk to glance and notify Ironic? | 13:05 |
| soren | Nova would be involved somehow, too. | 13:06 |
| soren | Nova keeps track of which kernel/ramdisk a given instance is running. | 13:06 |
| lucasagomes | right, I gotta take a better look into how nova manages it | 13:07 |
| lucasagomes | I wasn't aware of it | 13:08 |
| lucasagomes | I thought that all the references lived in glance | 13:08 |
| *** Marga_ has quit IRC | 13:09 | |
| *** Marga_ has joined #openstack-ironic | 13:09 | |
| soren | Wow, so I really have to do all of this manually? | 13:09 |
| * soren is baffled he's the first one to want to upgrade kernels on nodes managed by Ironic | 13:10 | |
| sambetts_ | lucasagomes: do you think that JayF might have some ideas on this? | 13:11 |
| lucasagomes | sambetts_, he might yeah | 13:12 |
| lucasagomes | sambetts_, tho IPA uses full disk images, so I bet that the kernel and ramdisk is a simple package update for that case | 13:13 |
| lucasagomes | soren, ^ yeah :/ | 13:13 |
| soren | Yeah, that's an entirely different story. | 13:13 |
| *** dlpartain has joined #openstack-ironic | 13:13 | |
| *** ryanpetrello has joined #openstack-ironic | 13:15 | |
| lucasagomes | soren, one thing I can say tho by looking at the Ironic code is that we always grab the kernel & ramdisk from the image metadata | 13:16 |
| lucasagomes | independent from what is on Nova's db | 13:16 |
| lucasagomes | https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/pxe.py#L232-L240 | 13:17 |
| soren | *headdesk* | 13:17 |
| * lucasagomes brb a sec, lunch | 13:17 | |
| lucasagomes | soren, yeah :/ | 13:17 |
| lucasagomes | soren, sorry for that... many of that logic came from nova bm too so I suspect even nova bm | 13:18 |
| soren | Uh, no? | 13:18 |
| soren | i_info is the instance info. | 13:18 |
| lucasagomes | might not be looking at the nova db either | 13:18 |
| lucasagomes | yup instance info | 13:18 |
| soren | If it *doesn't* have kernel and ramdisk already set, it grabs it from glance. | 13:18 |
| soren | So instance data takes precedence. As it should. | 13:18 |
| lucasagomes | yeah | 13:19 |
| lucasagomes | but nova doesn't populate that kernel ramdisk in instance_info afaict | 13:19 |
| lucasagomes | https://github.com/openstack/nova/blob/master/nova/virt/ironic/patcher.py | 13:19 |
| lucasagomes | it does for the deploy_ramdisk and deploy_kernel | 13:19 |
| lucasagomes | so, should be a easy fix in nova to look at the db and populate it correctly when the node is being deployed | 13:20 |
| * lucasagomes brb | 13:21 | |
| *** lucasagomes is now known as lucas-hungry | 13:21 | |
| *** bearhands is now known as comstud | 13:22 | |
| *** pensu has joined #openstack-ironic | 13:26 | |
| *** achanda has joined #openstack-ironic | 13:26 | |
| *** anderbubble has joined #openstack-ironic | 13:27 | |
| *** anderbubble has quit IRC | 13:30 | |
| *** anderbubble has joined #openstack-ironic | 13:30 | |
| *** romcheg has joined #openstack-ironic | 13:31 | |
| *** achanda has quit IRC | 13:31 | |
| *** romcheg has quit IRC | 13:33 | |
| *** romcheg has joined #openstack-ironic | 13:34 | |
| *** dlpartain has quit IRC | 13:37 | |
| naohirot | rameshg87: I really really missed precious opportunity to talk to you. | 13:39 |
| *** romcheg1 has joined #openstack-ironic | 13:39 | |
| naohirot | I just left to have dinner. | 13:39 |
| *** jjohnson2 has joined #openstack-ironic | 13:41 | |
| naohirot | I join to this channel everyday from today on. | 13:41 |
| *** romcheg2 has joined #openstack-ironic | 13:42 | |
| *** jjohnson2 has quit IRC | 13:42 | |
| *** romcheg has quit IRC | 13:43 | |
| *** jjohnson2 has joined #openstack-ironic | 13:43 | |
| *** igordcard has quit IRC | 13:43 | |
| *** romcheg1 has quit IRC | 13:43 | |
| naohirot | I can stay in this channel UTC 00:00-UTC14:00. | 13:44 |
| *** igordcard has joined #openstack-ironic | 13:44 | |
| *** romcheg has joined #openstack-ironic | 13:47 | |
| *** romcheg2 has quit IRC | 13:47 | |
| naohirot | rameshg87: I'm not familiar with IRC. Is there any way to show my status such Microsoft Lync? | 13:51 |
| naohirot | s/such/such as/ | 13:52 |
| *** lucas-hungry is now known as lucasagomes | 13:53 | |
| lucasagomes | naohirot, you can set ur away status | 13:53 |
| lucasagomes | like: /away <msg> | 13:54 |
| *** pensu has quit IRC | 13:54 | |
| lucasagomes | and /away with no message means ur no longer away :) | 13:54 |
| naohirot | Hi lucasagomes, again :-) | 13:54 |
| naohirot | Okay. | 13:54 |
| lucasagomes | or just change the nickname, people usually do like <nick>|<status>... like naohirot|sleeping | 13:55 |
| lucasagomes | :D | 13:55 |
| naohirot | Aha | 13:55 |
| naohirot | that's the reason some people change his/her name. | 13:55 |
| naohirot | lucasagomes: are you in Europe? | 13:56 |
| lucasagomes | naohirot, yup | 13:56 |
| naohirot | I see, PRIMERGY is productized in Germany. | 13:57 |
| naohirot | Can I ask basic question again? | 13:57 |
| naohirot | From user's point of view, | 13:58 |
| lucasagomes | sure | 13:58 |
| * lucasagomes googles primergy | 13:58 | |
| Shrews | lucasagomes: you want to address dtantsur's comments in https://review.openstack.org/129261 before I +A, or after? | 13:58 |
| lucasagomes | oh fujitsu :) | 13:58 |
| Shrews | lucasagomes: or at all? :) | 13:58 |
| lucasagomes | Shrews, morning Shrews | 13:58 |
| Shrews | morning | 13:59 |
| naohirot | GM, shrews | 13:59 |
| Shrews | hi naohirot | 13:59 |
| lucasagomes | Shrews, oh I can submit a new one if you think it worth fixing on that patch | 13:59 |
| naohirot | Is "nova boot" the only I/F user can see? | 13:59 |
| Shrews | lucasagomes: not necessary, i think. i'll +A and leave it up to you | 14:00 |
| lucasagomes | Shrews, ack! ta much! | 14:00 |
| lucasagomes | naohirot, the client has a ironic node-set-provision-state that can be used to start a deployment | 14:01 |
| lucasagomes | tho, we still want to make it simpler for the users to use Ironic without going through nova | 14:01 |
| naohirot | lucasagomes: I meant user as "end user", not "administrator". | 14:02 |
| lucasagomes | right now we have things used in the deployment like instance_uuid, which is nova-related only | 14:02 |
| lucasagomes | naohirot, oh gotcha | 14:02 |
| lucasagomes | yeah so nova is the only way | 14:02 |
| *** jistr|english is now known as jistr | 14:02 | |
| naohirot | Isn't it "ironic" for admin, right? | 14:02 |
| lucasagomes | its admin only nowadays | 14:02 |
| lucasagomes | some people wants to add policies but we haven't got any patches as far as I can tell | 14:03 |
| naohirot | lucasagomes: okay, in kilo cycle, is horizon team going to implement Web GUI for Ironic? | 14:03 |
| lucasagomes | JoshNang, was working on that since Juno | 14:03 |
| lucasagomes | but ofc in Juno it couldn't be merged cause Ironic wasn't even graduated | 14:04 |
| *** rushiagr is now known as rushiagr_away | 14:04 | |
| lucasagomes | but yeah, there are people working on that right now | 14:04 |
| naohirot | I see. | 14:04 |
| naohirot | And let me make suer if I understood the difference between pxe_ and agent_. | 14:05 |
| naohirot | In case of agent_, only the first boot is PXE, but the second boot and later are iSCSI. | 14:06 |
| naohirot | In case of pxe_, every time pxe boot, even though physical OS has been installed, because no ipl. | 14:07 |
| lucasagomes | nop | 14:07 |
| naohirot | really? | 14:07 |
| lucasagomes | agent_ deploys a full disk image, so first boot is PXE to actually boot the deploy ramdisk | 14:07 |
| lucasagomes | the rest is local boot | 14:07 |
| lucasagomes | in the pxe_ case, yes we pxe boot everytime | 14:08 |
| lucasagomes | the ISCSI is used by the pxe_ to be able to copy the image onto the disk | 14:08 |
| lucasagomes | so the pxe_ deploy ramdisk boots in the node, it exposes the disk via iscsi and pass it to the ironic conductor, the conductor them partition the iscsi disk and copy the partition image onto it | 14:09 |
| naohirot | Okay, now I think I fully understood the difference. | 14:09 |
| *** victor_lowther__ has quit IRC | 14:09 | |
| *** victor_lowther__ has joined #openstack-ironic | 14:09 | |
| lucasagomes | apart from documenting better and exposing those capabilities/differences via our API | 14:09 |
| *** victor_lowther__ has quit IRC | 14:09 | |
| lucasagomes | we also want to have only 1 deploy ramdisk | 14:10 |
| lucasagomes | instead one for pxe_ and one of agent_ | 14:10 |
| *** victor_lowther has joined #openstack-ironic | 14:10 | |
| lucasagomes | we may in a soon future merge everything into the agent ramdisk | 14:10 |
| naohirot | So I think that HBA and SAN storage works or not depends on the ramdisk. | 14:10 |
| lucasagomes | yeah | 14:10 |
| lucasagomes | the agent_ ramdisk actually has a service running on it | 14:10 |
| lucasagomes | so it's much more flexible and extensible | 14:10 |
| naohirot | In oder to create the ramdisk, | 14:11 |
| naohirot | the ways for agent_ and pxe_ are different, right? I believe. | 14:12 |
| naohirot | sorry | 14:12 |
| lucasagomes | yeah, and thats what need to be fixed before we merge both of them | 14:13 |
| lucasagomes | agent_ works with coreos where pxe_ with disk-image-builder | 14:13 |
| lucasagomes | there's a disk-image-builder patch upstream that NobodyCam is working on | 14:13 |
| lucasagomes | to be able to build the agent_ ramdisk | 14:13 |
| victor_lowther | hm... is there a spec out yet for the state machine changes we hashed out at the summit? | 14:13 |
| lucasagomes | and the agent_ ramdisk has some problems being tested on gate, memory constraint, we have to work with infra | 14:14 |
| lucasagomes | to solve that too (should be easy) | 14:14 |
| lucasagomes | victor_lowther, haven't seem one yet... but +1 to iron that out asap | 14:14 |
| lucasagomes | victor_lowther, mind putting one up? | 14:15 |
| victor_lowther | Maybe after I find the "how to write a spec" document. :) | 14:15 |
| *** datajerk has joined #openstack-ironic | 14:16 | |
| *** rloo has joined #openstack-ironic | 14:17 | |
| lucasagomes | victor_lowther, https://wiki.openstack.org/wiki/Ironic/Specs_Process :) | 14:17 |
| naohirot | lucasagomes: disk-image-builder creates ramdisk for pxe_, and IPA contains ramdisk for agent_. Am I right? | 14:17 |
| lucasagomes | I think that for the state machine we already agreed with the changes | 14:17 |
| lucasagomes | so we can fullfil a full spec | 14:18 |
| * Shrews interested in reading this spec | 14:18 | |
| lucasagomes | naohirot, IPA is the python service that runs in the agent ramdisk | 14:18 |
| lucasagomes | they use coreos tools to build their image | 14:18 |
| naohirot | does coreos tool have name? | 14:19 |
| lucasagomes | https://github.com/openstack/ironic-python-agent/tree/master/imagebuild/coreos | 14:19 |
| naohirot | Aha, now I'm getting better understanding :-) | 14:22 |
| *** rakesh_hs has quit IRC | 14:23 | |
| naohirot | lucasagomes: thaks a lot | 14:23 |
| naohirot | thanks a lot. | 14:23 |
| lucasagomes | naohirot, no problem at all :) | 14:24 |
| naohirot | So I had a really good progress today by having conversation with you rather than just reading several materials | 14:26 |
| naohirot | Now time to go to bed in Japan. | 14:27 |
| naohirot | see you tomorrow! | 14:27 |
| *** naohirot has quit IRC | 14:28 | |
| lucasagomes | victor_lowther, http://paste.openstack.org/show/131911/ in case u need a ASCII art of the state machine | 14:35 |
| lucasagomes | ignoring error states ofc | 14:35 |
| victor_lowther | lucasagomes: https://blueprints.launchpad.net/ironic/+spec/new-ironic-state-machine | 14:35 |
| *** romcheg has quit IRC | 14:35 | |
| lucasagomes | w00t! | 14:36 |
| *** andreykurilin has quit IRC | 14:36 | |
| * dtantsur is looking forward to this spec too | 14:36 | |
| victor_lowther | Now to write lots of probably-markdown-incompatible ascii art and confuse everyone who was not at the summit. :) | 14:36 |
| lucasagomes | lol | 14:37 |
| victor_lowther | I am following the process @ https://wiki.openstack.org/wiki/Blueprints, that more or less correct? | 14:38 |
| lucasagomes | yeah that's fine afaik, it's basically | 14:39 |
| victor_lowther | nm, I should read the backscroll. | 14:39 |
| lucasagomes | creates a bp in launchpad, and then fill out the spec template | 14:39 |
| * victor_lowther hunts down a rst major mode for emacs | 14:41 | |
| lucasagomes | victor_lowther, also for the spec, I think that everyone that was in the summit can contribute writing it | 14:41 |
| victor_lowther | ya | 14:41 |
| *** romcheg has joined #openstack-ironic | 14:41 | |
| lucasagomes | just need the initial kick :) | 14:41 |
| victor_lowther | That is the idea. | 14:42 |
| lucasagomes | nice :) | 14:42 |
| lucasagomes | I don't have anything fancy to write rst, but I do use a online viewer sometimes to test things out | 14:42 |
| lucasagomes | http://rst.ninjs.org/ | 14:42 |
| *** romcheg1 has joined #openstack-ironic | 14:43 | |
| lucasagomes | you can also check the syntax, with doc8 | 14:43 |
| lucasagomes | https://launchpad.net/doc8 | 14:43 |
| *** romcheg has quit IRC | 14:46 | |
| *** rushiagr_away is now known as rushiagr | 14:51 | |
| *** rushiagr is now known as rushiagr_away | 14:52 | |
| *** jjohnson2 has quit IRC | 14:53 | |
| *** jjohnson2_ has joined #openstack-ironic | 14:53 | |
| *** openstackgerrit has quit IRC | 14:55 | |
| *** rushiagr_away is now known as rushiagr | 15:05 | |
| *** pensu has joined #openstack-ironic | 15:07 | |
| *** romcheg1 has quit IRC | 15:10 | |
| lucasagomes | dtantsur, the only thing about the validate on the driver_vendor_passthru is that validate expects a task parameter | 15:12 |
| dtantsur | hmmm hmmm | 15:12 |
| lucasagomes | I don't have it on driver_vendor_passthru, I probabaly have to change th validate to accept a task or context | 15:12 |
| dtantsur | well yeah, as you don't have a node... | 15:12 |
| lucasagomes | that's why (assuming here) validate wasn't called for driver_vendor_passthru begore | 15:12 |
| lucasagomes | before* | 15:12 |
| lucasagomes | yeah | 15:12 |
| dtantsur | I don't know. let's talk after the call | 15:13 |
| lucasagomes | ok | 15:13 |
| *** foexle has quit IRC | 15:13 | |
| *** marcoemorais has joined #openstack-ironic | 15:19 | |
| *** marcoemorais1 has joined #openstack-ironic | 15:23 | |
| *** marcoemorais has quit IRC | 15:26 | |
| *** k4n0 has quit IRC | 15:27 | |
| Shrews | JayF: JoshNang: Just watched the recording of your decom talk. Great stuff. | 15:28 |
| pensu | Hi everyone, any idea where does devstack ironic store images? Need to delete some of those to free space on my VM! | 15:30 |
| *** foexle has joined #openstack-ironic | 15:32 | |
| Shrews | pensu: /opt/stack/data/ironic maybe? | 15:35 |
| lucasagomes | pensu, /opt/stack/devstack/files/ | 15:36 |
| lucasagomes | there's a images dir there | 15:37 |
| dtantsur | /opt/stack/data/ironic IIRC | 15:37 |
| pensu | okay, got it.....it's /opt/stack/devstack/files.....thanks lucasagomes! | 15:39 |
| pensu | dtantsur: Shrews: I don't have an /opt/stack/data/ironic folder.... | 15:39 |
| dtantsur | they must've changed something then... | 15:40 |
| lucasagomes | once you do a ./unstack.sh the /opt/stack/data/ironic goes away | 15:41 |
| *** pcrews has joined #openstack-ironic | 15:46 | |
| *** ryanpetrello has quit IRC | 15:48 | |
| *** achanda has joined #openstack-ironic | 15:59 | |
| *** ChuckC has quit IRC | 16:00 | |
| *** achanda has quit IRC | 16:07 | |
| *** ndipanov is now known as ndipanov_gone | 16:20 | |
| lucasagomes | dtantsur, alright so validate() | 16:29 |
| dtantsur | yeah | 16:29 |
| dtantsur | can we have something like driver_validate()? Or it doesn't make any sense? | 16:29 |
| * victor_lowther forsees lots of comment wars in the new state machine spec. | 16:30 | |
| lucasagomes | dtantsur, I don't know what would be best, 1- leave it as a known problem (TODO in the code), 2- modify the validate() to accept a task or context as the first parameter, 3- have a diff validate() endipoint for drivers | 16:30 |
| dtantsur | victor_lowther, oh be sure ;) | 16:30 |
| lucasagomes | dtantsur, yeah that would be 3- | 16:30 |
| JayF | sambetts_: lucasagomes: We use static DHCP (dhcp_provider None) with static agent images being served up, so I don't think I have much insight for you there :( | 16:30 |
| JayF | Shrews: ty | 16:30 |
| *** Marga_ has quit IRC | 16:31 | |
| lucasagomes | JayF, I'm a bit out of context... | 16:31 |
| dtantsur | lucasagomes, I'm inclined to go with 1- and unblock this work | 16:31 |
| victor_lowther | Actually writing it down is exposing lots of ambiguity. | 16:31 |
| JayF | lucasagomes: this is in response to something well, well overnight :) | 16:31 |
| *** Marga_ has joined #openstack-ironic | 16:31 | |
| dtantsur | victor_lowther, as it usually happens :) | 16:31 |
| JayF | lucasagomes: just catching up on scrollback :) | 16:31 |
| victor_lowther | Indeed. | 16:31 |
| JayF | lucasagomes: probably well past hte issue now | 16:31 |
| lucasagomes | JayF, ohh I remember now yeah it's about the full disk images | 16:31 |
| lucasagomes | gotcha | 16:31 |
| lucasagomes | dtantsur, right... yeah it didn't call validate() before too, so I will add a TODO | 16:32 |
| dtantsur | lucasagomes, so if you just leave a NOTE there, I'm ok with +2ing it | 16:32 |
| lucasagomes | I mean FIXME actually I can open a bug about it too | 16:32 |
| dtantsur | ack | 16:32 |
| lucasagomes | dtantsur, ack... yeah, even because that can be fixed in parallel | 16:32 |
| * lucasagomes doing it now | 16:32 | |
| *** marcoemorais1 has quit IRC | 16:34 | |
| lucasagomes | JayF, q about IPA... I think I know the answer but just confirming | 16:34 |
| *** ChuckC has joined #openstack-ironic | 16:34 | |
| lucasagomes | JayF, you guys don't install a bootloader right? The image already contains one | 16:34 |
| lucasagomes | is that correct? | 16:34 |
| JayF | Yes | 16:34 |
| lucasagomes | ok, is it grub or syslinux or ... ? | 16:34 |
| JayF | right now we just write a converted-to-raw image to the disk, and the image already has partitions/bootloaders/etc | 16:34 |
| JayF | depends on the image :) | 16:35 |
| *** ChuckC has quit IRC | 16:35 | |
| JayF | our CoreOS image uses syslinux, other images use grub | 16:35 |
| JayF | it's all up to the distribution | 16:35 |
| *** ChuckC has joined #openstack-ironic | 16:35 | |
| lucasagomes | right, just wondering because in case we want to add local boot support for the iSCSI methodology | 16:35 |
| lucasagomes | we may want to pick a bootloader | 16:35 |
| *** Marga_ has quit IRC | 16:35 | |
| JayF | In fact we rely on the image to have a reasonable partition table as well, and allow it to provide a configdrive partition for us to image into | 16:35 |
| JayF | if properly labelled | 16:35 |
| lucasagomes | and I know grub has some problems on being installed outside the machine that it's going to run | 16:36 |
| lucasagomes | so I think syslinux maybe the right choice | 16:36 |
| JayF | I mean, I don't think that holds true so much anymore | 16:36 |
| JayF | Just like with anything if you build an full disk image in one place (ours are built inside a VM iirc) you have to know things about your target to make it work | 16:37 |
| JayF | because things like the fstab, etc has to align | 16:37 |
| JayF | even beyond the bootloader | 16:37 |
| lucasagomes | indeed | 16:37 |
| dtantsur | going a bit earlier today, see you tomorrow | 16:39 |
| *** dtantsur is now known as dtantsur|afk | 16:39 | |
| lucasagomes | dtantsur|afk, see ya thanks | 16:40 |
| *** ChuckC has quit IRC | 16:50 | |
| lucasagomes | JayF, another quick q... for the gate problem with IPA | 16:52 |
| lucasagomes | are u guys looking into it? anything I can help with? | 16:52 |
| JayF | We've been playing whackamole with the various failures | 16:52 |
| JayF | last time it was an iptables rule blocking agent->swift | 16:53 |
| *** Viswanath has joined #openstack-ironic | 16:53 | |
| JayF | that was fixed just before the summit and none of us have looked since | 16:53 |
| JayF | afaik | 16:53 |
| lucasagomes | gotcha | 16:53 |
| lucasagomes | the memory thing has been sorted? | 16:53 |
| JayF | Yeah, basically IPA tests are forced to run with 1G in serial | 16:53 |
| JayF | until we can get IPA working in less memory (which requires DIB image and possibly image streaming support) | 16:54 |
| lucasagomes | gotcha | 16:54 |
| *** jistr has quit IRC | 16:55 | |
| JayF | Optimizing for low ram / slow connectivity was something far from our minds when designing this against hardware (back when it was teeth-agent before we integrated with Ironic) that had 32GB+ ram :) | 16:55 |
| JayF | heh | 16:55 |
| Shrews | all the things should be able to run in 256K | 16:56 |
| lucasagomes | heh yeah I don't think that it's a problem for the real world scenario indeed | 16:56 |
| *** Viswanath has quit IRC | 16:56 | |
| JayF | Shrews: I learned how to sysadmin about a decade ago fitting a whole lamp stack in one of those 32MB VPS boxes :) | 16:57 |
| JayF | Shrews: and here I thought my days of limited ram were behind me... :P | 16:57 |
| lucasagomes | :D | 16:58 |
| Shrews | JayF: welcome to the cloud of the 90's! :) | 16:59 |
| *** ChuckC has joined #openstack-ironic | 16:59 | |
| *** leopoldj has quit IRC | 17:03 | |
| lucasagomes | +1 /me hates the term cloud | 17:04 |
| *** marcoemorais has joined #openstack-ironic | 17:13 | |
| *** dlaube has joined #openstack-ironic | 17:20 | |
| *** athomas has quit IRC | 17:20 | |
| *** andreykurilin has joined #openstack-ironic | 17:22 | |
| *** rameshg87 has joined #openstack-ironic | 17:24 | |
| *** Marga_ has joined #openstack-ironic | 17:25 | |
| * sambetts_ agrees with lucasagomes | 17:31 | |
| *** ifarkas has quit IRC | 17:33 | |
| *** penick has joined #openstack-ironic | 17:34 | |
| *** ifarkas has joined #openstack-ironic | 17:39 | |
| *** Marga_ has quit IRC | 17:42 | |
| yjiang5 | lucasagomes: I noticed your statement "in-band will be a separated service". Do you have any etherpad for the discussion? | 17:44 |
| *** derekh has quit IRC | 17:44 | |
| *** openstackgerrit has joined #openstack-ironic | 17:49 | |
| *** pelix has quit IRC | 17:53 | |
| *** foexle has quit IRC | 17:56 | |
| openstackgerrit | A change was merged to openstack/ironic: Fix documentation on Standard driver interfaces https://review.openstack.org/133744 | 17:58 |
| *** anderbubble has joined #openstack-ironic | 17:59 | |
| JoshNang | Shrews: thanks :D | 17:59 |
| *** ChuckC has quit IRC | 18:00 | |
| openstackgerrit | Ruby Loo proposed a change to openstack/ironic-specs: Updates to 'maintenance reason' spec https://review.openstack.org/133776 | 18:03 |
| rloo | hi ironickers | 18:03 |
| Shrews | hi rloo. welcome back | 18:04 |
| rloo | lucasagomes, jroll: I updated the maintenance reason spec ^^ 133776 to have an APIImpact flag in the commit, so I had to change something ;) You might want to check that I got it right. | 18:04 |
| rloo | hi Shrews, we missed you! | 18:05 |
| lucasagomes | rloo, morning will check | 18:05 |
| lucasagomes | rloo, btw I also have to talk to u about the vendor spec | 18:05 |
| rloo | lucasagomes: ok. am still catching up on stuff. is there something you want me to look at first, wrt vendor spec? | 18:05 |
| lucasagomes | rloo, remember we talked about doing a GET on the vendor_passthru root endpoint to get the avaialble methods? | 18:05 |
| lucasagomes | rloo, well not really, just telling you | 18:06 |
| rloo | lucasagomes: yes wrt the GET | 18:06 |
| lucasagomes | rloo, yup, so that doesn't work | 18:06 |
| rloo | hmm, too bad. | 18:06 |
| lucasagomes | cause the methods for the vendor passthru is an argument | 18:06 |
| lucasagomes | like {'method': blah} .../vendor_passthru OR ../vendor_passthru/blah | 18:06 |
| rloo | lucasagomes: but aren't the methods mapped in the dict? | 18:06 |
| lucasagomes | so we already use GET for the root | 18:06 |
| lucasagomes | you can do both | 18:07 |
| rloo | oh. | 18:07 |
| lucasagomes | it's just a syntax thing, in my patch I will have to add a new endpoint | 18:07 |
| lucasagomes | like .../vendor_passthru/methods or so | 18:07 |
| rloo | that's fine, adding a new endpoint. | 18:07 |
| lucasagomes | cool, yeah just to say because it will differ from the spec a bit :/ | 18:07 |
| rloo | I think we might want to update specs after they are approved, if we find that they need updating. I think it was mentioned (maybe in nova?) that they do this. | 18:08 |
| rloo | or maybe it was a question as to whether they should be updated after they are approved. I don't recall, but I'm changing it in my reality to yes, we should. ha ha. what do people think? | 18:08 |
| lucasagomes | +2 ur patch, looks good more clear | 18:09 |
| lucasagomes | I think it's OK to update the specs after approved | 18:09 |
| lucasagomes | I don't want someone to know everything when writing the spec | 18:10 |
| lucasagomes | many things we forget and ended up remembering/learning as we code | 18:10 |
| lucasagomes | so I think it's complete fine to update a merged spec | 18:10 |
| rloo | lucasagomes: great, glad you agree too. | 18:10 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/python-ironicclient: Sane parameters for node and driver vendor_passthru() https://review.openstack.org/133780 | 18:13 |
| rloo | lucasagomes: so, ahh, would you mind updating the 'Extend Vendor Passthru' to mention the new endpoint? It needs some patch that has an APIImpact flag and I was going to find something to patch, but since you're going to change it. Or have you decided what to use, 'methods'? I can update it for you. | 18:13 |
| lucasagomes | rloo, +1 I will probably do it tomorrow cause it's a bit late here | 18:13 |
| lucasagomes | but yeah I will update it | 18:14 |
| lucasagomes | yeah on my patch I added a "methods" endpoint | 18:14 |
| lucasagomes | couldn't think of a better name | 18:14 |
| rloo | lucasagomes: no worries. I'm not in a hurry, just want to make sure any approved specs that should have had APIImpact, has it ;). thx. | 18:14 |
| lucasagomes | GET vendor_passthru/methods | 18:14 |
| rloo | lucasagomes: I think 'methods' is fine, since that's what we're calling them, right? | 18:14 |
| lucasagomes | would return all the methods supported by the driver/node | 18:14 |
| *** athomas has joined #openstack-ironic | 18:14 | |
| lucasagomes | yeah | 18:14 |
| lucasagomes | rloo, aight! adding to the TODO list for tomorrow then | 18:15 |
| rloo | thx lucasagomes | 18:15 |
| *** achanda has joined #openstack-ironic | 18:17 | |
| *** mjturek has left #openstack-ironic | 18:17 | |
| rameshg87 | JayF, hi | 18:18 |
| lucasagomes | alright I will call it a day | 18:18 |
| lucasagomes | have a good night everyone! | 18:18 |
| rameshg87 | goodnight lucasagomes :) | 18:18 |
| sambetts_ | same here, good night ironic | 18:19 |
| lucasagomes | sambetts_, night | 18:19 |
| lucasagomes | rameshg87, thanks you too :) | 18:19 |
| sambetts_ | lucasagomes: night! | 18:19 |
| *** lucasagomes is now known as lucas-zZz | 18:19 | |
| *** sambetts_ has quit IRC | 18:19 | |
| *** spandhe has joined #openstack-ironic | 18:20 | |
| JayF | rameshg87: hey | 18:21 |
| rameshg87 | JayF, hello | 18:21 |
| rameshg87 | JayF, ifarkas and myself were looking for you | 18:22 |
| rameshg87 | JayF, we were thinking of coming up with a common approach of doing raid configuration for dell and hp servers | 18:22 |
| *** anderbubble has quit IRC | 18:22 | |
| JayF | rameshg87: for the out of band stuff I don't have much experience or really a dog in the fight at all :) | 18:23 |
| rameshg87 | JayF, are you planning to expand this spec: https://review.openstack.org/#/c/131272/1/specs/kilo/hardware-capabilities.rst ? | 18:23 |
| rameshg87 | JayF, yeah that's fine, hp is targetting inband :) | 18:23 |
| JayF | rameshg87: I'd love for someone to take over that spec and reflect all the work done in the design summit session | 18:23 |
| rameshg87 | JayF, i could give a try then | 18:24 |
| JayF | please do | 18:24 |
| rameshg87 | JayF, i might need to work with dtantsur|afk so that we come up with a common approach of putting together driver capabilities and hardware capabilities | 18:24 |
| JayF | I think that's the right path | 18:24 |
| rameshg87 | JayF, i don't see a reason why those two should be separate | 18:24 |
| rameshg87 | JayF, for hp, we are giving a try putting the hp raid configuration tool from the ipa image | 18:25 |
| JayF | There's a different angle of capabilities that you probably aren't thinking about | 18:25 |
| JayF | i.e. as a user I have no way of knowing that pxe driver doesn't do whole disk images and ipa does | 18:25 |
| rameshg87 | JayF, what's that ? | 18:25 |
| JayF | I think driver capabilities is more targetting documenting those differences | 18:25 |
| JayF | than driving configuration like the node/hw capabilities stuff is | 18:25 |
| rameshg87 | JayF, but we decided disk vs partition images thing will be fixed | 18:26 |
| rameshg87 | JayF, is there any other use-case for driver capability ? | 18:26 |
| JayF | "will be" is how long from now? | 18:27 |
| *** Marga_ has joined #openstack-ironic | 18:27 | |
| JayF | and are you sure all deploy drivers will always support the same set of things? | 18:27 |
| JayF | I'm not convinced that's a technical possibility | 18:27 |
| rameshg87 | JayF, then it's merely not just documentation. driver capability is an actual functionality which can affect scheduling | 18:28 |
| JayF | that's true; but it should also be separately discoverable via the api | 18:28 |
| rameshg87 | JayF, and published to the nova scheduler | 18:28 |
| rameshg87 | JayF, or rather all set(hardware capabilities, driver capabilities) can be just termed as "capabilities" for the node | 18:29 |
| JayF | Yes, but the driver caps need to be separately discoverable imo | 18:29 |
| rameshg87 | JayF, why is that ? for bettering user understanding ? | 18:29 |
| JayF | Yes | 18:30 |
| rameshg87 | JayF, okay | 18:30 |
| rameshg87 | JayF, just one more question | 18:30 |
| JayF | user/operators have *no way whatsoever* to divine that knowledge today | 18:30 |
| JayF | unless they go diving into python | 18:30 |
| *** lucas-zZz has quit IRC | 18:30 | |
| rameshg87 | JayF, during the summit i heard something like certain things cant be done in ipa because ipa runs in docker container | 18:31 |
| *** Marga_ has quit IRC | 18:31 | |
| rameshg87 | JayF, things dealing with devices, is that true ? | 18:31 |
| *** jlk has quit IRC | 18:31 | |
| JayF | That's very not true | 18:32 |
| JayF | we run IPA inside a systemd-nspawn container with the --share-system flag | 18:32 |
| JayF | so it acts more like a chroot than a container | 18:32 |
| JayF | and IPA has full hardware access | 18:32 |
| *** krtaylor has quit IRC | 18:32 | |
| rameshg87 | JayF, okay and all device files in /dev | 18:33 |
| *** foexle has joined #openstack-ironic | 18:33 | |
| JayF | yeah | 18:33 |
| JayF | everything should work exactly the same as you'd expect | 18:33 |
| JayF | any differences would be because CoreOS runs a newer kernel or are a bug against the image | 18:33 |
| rameshg87 | JayF, okay :).. that's all .. let me try to polish the hardware capabilities spec .. | 18:34 |
| *** anderbubble has joined #openstack-ironic | 18:37 | |
| *** rameshg87 has quit IRC | 18:41 | |
| *** ChuckC has joined #openstack-ironic | 18:42 | |
| *** rushiagr is now known as rushiagr_away | 18:43 | |
| yjiang5 | Can anyone have a look on https://review.openstack.org/#/c/132936/ ?It's waiting some input from ironic guys. | 18:44 |
| *** ChuckC has quit IRC | 18:47 | |
| spandhe | Hi everyone.. observed something with Icehouse version of Ironic and wanted to confirm if thats expected behavior | 18:52 |
| spandhe | I have setup a tftp server, and it serves from a particular directory | 18:52 |
| spandhe | so all the paths that the target uses, e.g. pxelinux.0 or pxelinux.cfg/ etc, are related to that path | 18:53 |
| spandhe | relative to* | 18:53 |
| spandhe | but when Ironic creates the pxe config file, it puts absolute paths for kernel and ramdisk images | 18:54 |
| spandhe | so pxe fails to find those files.. | 18:54 |
| spandhe | is this expected? | 18:55 |
| spandhe | I mean, Its not expected to fail, but are the paths in the pxe config file supposed to be absolute? | 18:55 |
| spandhe | I am serving those images through tftp also.. | 18:55 |
| *** igordcard has quit IRC | 18:55 | |
| *** krtaylor has joined #openstack-ironic | 18:56 | |
| *** marcoemorais has quit IRC | 18:58 | |
| *** marcoemorais has joined #openstack-ironic | 18:58 | |
| *** marcoemorais has quit IRC | 18:59 | |
| *** marcoemorais has joined #openstack-ironic | 18:59 | |
| *** achanda has quit IRC | 19:00 | |
| yjiang5 | JayF: Hi, thanks for reviewing https://review.openstack.org/#/c/132936/3/lib/ironic . You mean I should not adding the agent_ipmitool to the IRONIC_ENABLED_DRIVERS (as in https://review.openstack.org/#/c/132936/3/lib/ironic L58 ), or the whole patch? | 19:02 |
| JoshNang | yjiang5: JayF i think we should add it. It's an in-tree driver, we should do what we can to make it a first class, fully tested citizen in Kilo | 19:06 |
| JayF | I honestly want devananda to weigh in | 19:07 |
| JayF | because I was under the impression only the default drivers were enabled by default in devstack | 19:07 |
| JayF | i.e. a breaking change to agent driver shouldn't break nova's (or anyone elses' gate, except Ironic) gate | 19:08 |
| spandhe | rloo: hey.. do u know if thats a bug? | 19:10 |
| yjiang5 | JayF: so you mean we should not add this patch to devstack, right? | 19:12 |
| rloo | spandhe: hi, sorry, you mean your question about tftp dir? (wasn't paying attention) | 19:12 |
| spandhe | rloo: yes.. about tftp directory | 19:12 |
| JayF | yjiang5: I'm saying I don't think the default devstack configuration should contain non-default Ironic drivers. | 19:12 |
| *** pensu has quit IRC | 19:13 | |
| yjiang5 | JayF: but we have agent_ssh driver in it alrleady. | 19:13 |
| rloo | spandhe: not sure. I'm not familiar with that code. but based on your description, it is a bug cuz it doesn't work for you? | 19:13 |
| devananda | JayF: you rang? | 19:13 |
| rloo | spandhe: did you try looking at the reported bugs and/or what was fixed/addressed in juno? | 19:14 |
| JayF | devananda: There's a review up for devstack that yjiang5 made that would enable agent_ipmitool in default ironic devstack config | 19:14 |
| JayF | devananda: iirc at the mid-cycle in Oregon we said that shouldn't happen unless/until it was default driver? | 19:14 |
| devananda | JayF: i also recall that being the thing we said | 19:15 |
| devananda | defaults in devstack != recommendations for production | 19:15 |
| JayF | devananda: wanted to make sure I wasn't misremembering :) this is the merge req in question --> https://review.openstack.org/#/c/132936/ | 19:15 |
| *** Marga_ has joined #openstack-ironic | 19:15 | |
| devananda | our config option should resemble something production-sane | 19:16 |
| devananda | devstack's defaults should resemble something useful in a devstack environment ... | 19:16 |
| devananda | that's jus tmy 2c, though | 19:16 |
| spandhe | rloo: it doesnt work for me, yes.. but now I am wondering if it assumes that some http service is supposed to serve the file instead of tftp.. or whether I am doing something wrong thats causing absolute paths to show up there | 19:16 |
| yjiang5 | JayF: devananda, sorry guys, what do you mean the default driver? | 19:16 |
| spandhe | if more people have observed this issue, its probably a bug.. | 19:16 |
| JayF | yjiang5: default deploy driver for ironic right now is "pxe" not "agent" | 19:16 |
| rloo | spandhe: it would be easier if you were using juno | 19:17 |
| rloo | spandhe: 'easier' == a lot was changed | 19:17 |
| devananda | JayF: devstack currently enables agent_ssh, pxe_ssh, and pxe_ipmitool | 19:17 |
| spandhe | rloo: :) hopefully soon :) | 19:17 |
| devananda | this is closing a gap in that coverage matrix | 19:17 |
| yjiang5 | JayF: devananda So you mean agent_ssh in not default driver although it's supported in devstack, right? I'm a bit lost. | 19:17 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties using new CLI node-discover-properties https://review.openstack.org/100951 | 19:17 |
| JayF | devananda: ooooooh. So the big difference here is enabling ipmitool | 19:18 |
| devananda | right | 19:18 |
| devananda | ipmitool + agent is missing from the *default* list of enabled drivers in devstack | 19:18 |
| *** achanda has joined #openstack-ironic | 19:18 | |
| devananda | it's totally possible to enable it by overriding IRONIC_ENABLED_DRIVERS in your localrc, of course | 19:18 |
| devananda | if we're not exercisign it in the gate, https://review.openstack.org/#/c/132936/3/lib/ironic L58 isn't helpful | 19:19 |
| devananda | however ... the other changes iin that file should land regardless | 19:19 |
| Shrews | agent_ssh is for the IPA tests, iirc. i see no need to add the ipmitool one "just because" | 19:19 |
| devananda | there's a bug right now in devstack's handling, it seems. using agent_[!ssh] is not possible without modifying this file | 19:19 |
| devananda | look at the change on line 497 | 19:20 |
| devananda | that definitely should land | 19:20 |
| devananda | ditto L 294 | 19:20 |
| yjiang5 | devananda: Sure, I will remove that line. that's my question to JayF :) | 19:20 |
| JayF | if you drop your addition of agent_ipmitool to the IRONIC_ENABLED_DRIVERS I'll take my -1 off :) | 19:21 |
| devananda | though.... [[ ]] shouldn't be used. we can do better pattern matching than =~ agent_[a-z] with no quotes | 19:21 |
| JayF | assuming that's the general consensus of the channel as well | 19:21 |
| rloo | spandhe: I looked quickly at git log and I see a few changes related to tftp. It is possible that there was a bug in icehouse that is now fixed. w/o looking through the code though, I don't know definitively. | 19:21 |
| yjiang5 | Shrews: I'm not sure. In fact, I'm using devstack for my mini-cloud and that's the reason I added the pxe_ipmitool (https://review.openstack.org/#/c/127658/6 ) and agent_ipmitool (this patch) :) | 19:21 |
| * devananda adds comments | 19:21 | |
| *** marcoemorais has quit IRC | 19:21 | |
| spandhe | rloo: thanks.. Ill check Juno code.. | 19:21 |
| *** marcoemorais has joined #openstack-ironic | 19:22 | |
| Shrews | yjiang5: like devananda said, those can be enabled in local.conf | 19:22 |
| devananda | also, I'll be around on weird schedules for this week | 19:22 |
| yjiang5 | devananda: sorry, what do you mean of "we can do better pattern matching than =~ agent_[a-z] with no quotes". | 19:22 |
| * Shrews wonders why pxe_ipmitool is enabled by default | 19:23 | |
| yjiang5 | Shrews: yes, I hesitated to that changes also when submitting the patch, because agent_ssh is not there, just didn't remove it. :) | 19:23 |
| Shrews | i wonder if d-g overrides this value... | 19:23 |
| * Shrews looks | 19:23 | |
| devananda | Shrews: I wonder that too... | 19:23 |
| yjiang5 | Shrews: possibly because pxe_ipmitool is in the default driver already. | 19:23 |
| Shrews | devananda: https://github.com/openstack-infra/devstack-gate/blob/master/devstack-vm-gate.sh#L139-L162 | 19:25 |
| Shrews | only for the agent tests | 19:26 |
| devananda | Shrews: yup. so there's no need for it to be in the defaults right now. but also no need to remove it | 19:27 |
| devananda | i mean, we could clean that up a bit | 19:27 |
| Shrews | devananda: right | 19:27 |
| Shrews | but definitely no need to add something else :) | 19:28 |
| yjiang5 | devananda: for "better pattern matching than =~ agent_[a-z] ", do you have any suggestion of better method? Sorry not well on bash . | 19:30 |
| devananda | yjiang5: if [ -z "${IRONIC_DEPLOY_DRIVER%%agent*}" ]; then | 19:34 |
| devananda | tests if the variable begins with the word "agent" | 19:34 |
| yjiang5 | devananda: thanks, I will try that way. | 19:34 |
| yjiang5 | devananda: I will update the previous patch also, which add pxe_ipmitool support. | 19:35 |
| soren | devananda: Uh, no. That checks if it contains the word "agent". | 19:36 |
| devananda | soren: -z makes it check if it starts with it | 19:36 |
| soren | foo=baragentbaz; echo ${foo%%agent*} gives "bar". | 19:36 |
| devananda | soren: correct. with -z, checks for the string to be empty | 19:37 |
| soren | devananda: Yes... | 19:37 |
| soren | And it's not. | 19:37 |
| soren | Uh... | 19:37 |
| soren | Sheesh. | 19:37 |
| soren | I'm an idiot. | 19:37 |
| devananda | I am too | 19:37 |
| devananda | but this time, I tested it :) | 19:37 |
| soren | You're right. It checks if the string begins with "agent" or is the empty string. | 19:38 |
| *** igordcard has joined #openstack-ironic | 19:38 | |
| soren | ...but the latter case is probably irrelevant. | 19:38 |
| soren | I always get confused by negations. I'm clearly not in the line of business. | 19:38 |
| yjiang5 | soren, thanks for double check and confirmation. | 19:39 |
| soren | In the /right/ line of business. | 19:39 |
| devananda | yjiang5: comments posted. other than the nits, the patch LGTM | 19:39 |
| soren | (See? I can't even type properly) | 19:39 |
| yjiang5 | devananda: thanks. | 19:39 |
| * devananda wonders how late his plane is going to be | 19:39 | |
| devananda | oh hey, anyone using reveal.js notes-server functionality? | 19:40 |
| soren | Now that there are more folks here... Does anyone have a secret trick for upgrading kernels on nodes managed by Ironic? (pxe/ipmi driver) | 19:40 |
| *** anderbubble has quit IRC | 19:41 | |
| yjiang5 | soren: switch to IPA, that's what I'm doing now :) | 19:42 |
| devananda | oh great. my flight to dallas is delayed enough that i'm likely to miss my connection now. /me goes afk to find an agent to fix this | 19:42 |
| *** anderbubble has joined #openstack-ironic | 19:42 | |
| devananda | soren: change the kernel/ramdisk ref in node.properties and reboot? | 19:42 |
| devananda | ah. no. that SHOULD work but doesn't | 19:42 |
| soren | devananda: "node.properties" ? | 19:42 |
| soren | Mind you, my Ironic install is a bit long in the tooth. | 19:43 |
| soren | Like, ice house release or something. | 19:43 |
| yjiang5 | soren: oooook :) | 19:43 |
| devananda | soren: right. the same thing should (not work) there because i dont think that code has changed much | 19:43 |
| yjiang5 | soren: You can simiply replace the the file in the local cache? | 19:43 |
| devananda | but I don't think the pxe_ipmitool driver will rebuild the PXE env, even if you change the right property | 19:44 |
| devananda | soren: I'd call that a bug -- mind filing one? you *should* be able to update the instance's kernel & ramdisk and trigger a reboot | 19:44 |
| soren | devananda: Yeah, as far as I can tell, the .prepare method only gets called in response to a new deploy. | 19:44 |
| devananda | that is possible today with a "rebuild" | 19:44 |
| devananda | but that'll erase any data on the instance (unless you had an ephemeral volume and did "rebuild --preserve-ephemeral" ... don't ask ...) | 19:44 |
| soren | devananda: Yeah, I had that chat with Spamaps last week. :) | 19:45 |
| devananda | soren: or a rebuild, or a hash repartitioning event (in Juno) | 19:45 |
| devananda | anyhow, i need to sort out this flight prob lem | 19:45 |
| devananda | bbiab (or not) | 19:45 |
| soren | devananda: Enjoy | 19:45 |
| yjiang5 | Does anyone know if the key_name parameter is working for ironic? A quick checking of the code seems no. | 19:45 |
| *** igordcard has quit IRC | 19:52 | |
| *** penick has quit IRC | 19:52 | |
| *** shakamun_ has quit IRC | 19:52 | |
| *** marcoemorais has quit IRC | 19:53 | |
| *** marcoemorais has joined #openstack-ironic | 19:53 | |
| *** shakamunyi has joined #openstack-ironic | 19:57 | |
| *** Marga_ has quit IRC | 20:06 | |
| *** foexle has quit IRC | 20:06 | |
| *** anderbubble has quit IRC | 20:09 | |
| *** killer_prince has joined #openstack-ironic | 20:09 | |
| *** killer_prince is now known as lazy_prince | 20:09 | |
| *** ChuckC___ has joined #openstack-ironic | 20:18 | |
| *** ChuckC___ has quit IRC | 20:25 | |
| *** jrist has quit IRC | 20:28 | |
| *** penick has joined #openstack-ironic | 20:34 | |
| openstackgerrit | Victor Lowther proposed a change to openstack/ironic-specs: New Ironic provisioner state machine. https://review.openstack.org/133828 | 20:34 |
| victor_lowther | Let the reviewinating begin! | 20:36 |
| * Shrews eagerly notes the spec needing reviewification | 20:46 | |
| *** achanda has quit IRC | 20:46 | |
| JoshNang | victor_lowther: thanks for doing that spec. i thought i was going to have to :) | 20:54 |
| *** ifarkas has quit IRC | 20:55 | |
| Shrews | hrm, wonder if we really need individual fail states, or can just glom on to a single one | 20:56 |
| JayF | The individual states are WONDERFUL | 20:57 |
| *** krtaylor has quit IRC | 20:57 | |
| JayF | We key on them all the time for internal dashboards and scripts | 20:58 |
| Shrews | JayF: you say "them", but we currently only have a single fail state | 20:59 |
| JoshNang | you really want to treat ZAPFAIL different than DEPLOYFAIL or ERROR | 21:00 |
| JayF | Error + Deploy Failed upstream atm afik, right? | 21:00 |
| JayF | We added Decom Failed downstream | 21:00 |
| Shrews | JoshNang: of course, but the logic handling the error should know if it was a ZAP or DEPLOY that failed | 21:02 |
| JayF | the logic that happens once it hits that state is explicitly "operator intervention" | 21:02 |
| JayF | my operational tools should be able to know at a glance what state things are in | 21:02 |
| JayF | and a generic ERROR state that doesn't indicate what step it failed in or how it failed isn't that useful, especially considering how wildly different our error messages are | 21:03 |
| Shrews | fail is fail. perhaps you want to know from the state WHAT caused the failure though | 21:03 |
| Shrews | so, i could see *that* | 21:03 |
| JayF | fail is not fail | 21:03 |
| JayF | why and what was happening when fail happened matters a lot | 21:03 |
| Shrews | JayF: please tell that to my high school teachers | 21:04 |
| JayF | especially at scale when you're trying to automatically recover machines from bad states | 21:04 |
| JayF | this is all about wrapping Ironic in automation to recover the failed states | 21:04 |
| *** penick has quit IRC | 21:04 | |
| JayF | and the state being as specific as possible as to what and when it failed is extra input to that goal | 21:04 |
| mrda | Morning Ironic! | 21:05 |
| Shrews | JayF: i shall rephrase: if you're saying an *external* entity needs to know what failed, then i can see individual states being useful | 21:05 |
| JayF | Shrews: absolutely | 21:05 |
| JayF | Shrews: my bucket full of scripts that fix boxes in decom failed /deploy failed / etc | 21:05 |
| Shrews | i was considering from internal within ironic, at first | 21:06 |
| JayF | Shrews: plus we have a dashboard that interrogates the API for things like that :) This is absolutely an external integratino helper | 21:06 |
| JayF | since inherently the *FAIL states are "operator intervention required" | 21:06 |
| Shrews | JayF: ok ok. i'm convinced :-P | 21:07 |
| JoshNang | mrda: g'morning! | 21:07 |
| Shrews | mrda: morning | 21:07 |
| JayF | Shrews: never argue with someone who can type faster than you ;) | 21:07 |
| Shrews | JayF: maybe i need a dvorak keyboard to keep up | 21:09 |
| *** Viswanath has joined #openstack-ironic | 21:09 | |
| JayF | Shrews: I'm just kidding of course :P | 21:10 |
| mrda | \o | 21:10 |
| Shrews | personally, i can't wait for robot-monkey helpers to type for me | 21:10 |
| *** Marga_ has joined #openstack-ironic | 21:11 | |
| mrda | be careful, they will be robot monkey overlords | 21:11 |
| Shrews | I, for one, welcome our poo-flinging robotic overlords | 21:12 |
| *** Viswanath has quit IRC | 21:12 | |
| *** achanda has joined #openstack-ironic | 21:14 | |
| victor_lowther | individual fail states are doubleplusgood. | 21:14 |
| *** penick has joined #openstack-ironic | 21:16 | |
| victor_lowther | the recovery steps you probably want to take for DELETEFAIL (mostly automated, probably ignorable) are not the same ones you want for ZAPFAIL (OMG firmware flash hang! Call Support!) | 21:16 |
| Shrews | victor_lowther: please see the above scrollback wherein I yielded to JayF's superior intellect | 21:17 |
| Shrews | :) | 21:17 |
| victor_lowther | damn you scrollback! | 21:18 |
| JayF | victor_lowther: in most of my downstream experience, DECOMFAIL ("ZAPFAIL") correllates strongly to hardware issues, whereas DEPLOYFAIL (which downstream is both deploy filed and delete failed) usually indicates some kind of epehemeral brokenness (like an API that's flaky, or issues downloading the image, etc) | 21:18 |
| *** shakamunyi has quit IRC | 21:18 | |
| victor_lowther | ya | 21:18 |
| victor_lowther | I seperated DEPLOYFAIL and DELETEFAIL because consistency | 21:19 |
| victor_lowther | and DEPLOYFAIL can mean "Why did the BIOS not switch boot modes and/or flip the VT bit"? | 21:19 |
| victor_lowther | which probably also require manual intervention beyond a simple retry. | 21:20 |
| JayF | I guess so... I feel like DEPLOYING is overloaded a little bit | 21:20 |
| *** anderbubble has joined #openstack-ironic | 21:20 | |
| JayF | but right now it's very, very overloaded so this is still better | 21:20 |
| victor_lowther | that is why I had srt on the board at the summit | 21:20 |
| victor_lowther | but the majority of the room was fine with doing that stuff in deploying. | 21:21 |
| JayF | We just need to do a good job following up after this of standardizing error messages | 21:21 |
| JayF | because it should be easy to tell the difference between deployfail (prep step) and deployfail (imaging/primary deploy stuff failed) | 21:22 |
| Shrews | JayF: are you talking about parsing error messages for that differentiation? | 21:25 |
| JayF | Shrews: we absolutely parse error messages downstream today in our scripts | 21:26 |
| Shrews | ick | 21:26 |
| JayF | Shrews: which is why I'd love for one day for us to standardize on sets of error messages | 21:26 |
| Shrews | JayF: perhaps adding something like a "reason code" field would be better | 21:26 |
| JoshNang | ick is right, but it works pretty well | 21:26 |
| JayF | Shrews: imagine it this way: we classify servers as having given "moods" based on combinations of info given back from the API, including state, last_error, etc | 21:26 |
| JayF | Shrews: once you determine the ~5 most common ways things fail, matching on the error isn't so bad | 21:27 |
| Shrews | i see. but it's horrible to imagine that changing an error string upstream might break someone downstream. a numeric code would help with that | 21:28 |
| Shrews | but... hindsight and all | 21:28 |
| *** Marga__ has joined #openstack-ironic | 21:33 | |
| JayF | Shrews: I wouldn't consider that "breaking" me downstream fwiw, I'd just see new unknown statuses in my dashboard and fix my classification :) | 21:33 |
| *** shakamunyi has joined #openstack-ironic | 21:34 | |
| *** Marga_ has quit IRC | 21:35 | |
| *** ryanpetrello has joined #openstack-ironic | 21:36 | |
| *** Marga__ has quit IRC | 21:38 | |
| *** Marga_ has joined #openstack-ironic | 21:39 | |
| *** shakamunyi has quit IRC | 21:39 | |
| *** krtaylor has joined #openstack-ironic | 21:43 | |
| *** penick has quit IRC | 21:44 | |
| *** marcoemorais has quit IRC | 21:52 | |
| *** marcoemorais has joined #openstack-ironic | 21:53 | |
| *** ChuckC_ has joined #openstack-ironic | 21:54 | |
| *** ChuckC_ is now known as ChuckC | 21:54 | |
| *** penick has joined #openstack-ironic | 21:56 | |
| *** Marga_ has quit IRC | 22:04 | |
| *** marzif has joined #openstack-ironic | 22:05 | |
| *** ChuckC has quit IRC | 22:07 | |
| *** ChuckC has joined #openstack-ironic | 22:07 | |
| *** ryanpetrello has quit IRC | 22:07 | |
| *** jjohnson2_ has quit IRC | 22:11 | |
| openstackgerrit | A change was merged to openstack/ironic: Support ipxe with Dnsmasq https://review.openstack.org/130492 | 22:12 |
| *** yjiang5 is now known as yjiang5_away | 22:13 | |
| *** anderbubble has quit IRC | 22:33 | |
| *** ChuckC has quit IRC | 22:39 | |
| *** ChuckC has joined #openstack-ironic | 22:44 | |
| *** marzif has quit IRC | 22:46 | |
| *** anderbubble has joined #openstack-ironic | 22:46 | |
| *** rushiagr_away is now known as rushiagr | 22:48 | |
| *** jrist has joined #openstack-ironic | 22:51 | |
| *** shakamunyi has joined #openstack-ironic | 22:59 | |
| *** jrist has quit IRC | 22:59 | |
| *** jrist has joined #openstack-ironic | 23:00 | |
| *** shakamun_ has joined #openstack-ironic | 23:00 | |
| *** shakamunyi has quit IRC | 23:04 | |
| *** athomas has quit IRC | 23:05 | |
| *** yuanying has joined #openstack-ironic | 23:07 | |
| *** shakamunyi has joined #openstack-ironic | 23:07 | |
| *** shakamun_ has quit IRC | 23:11 | |
| *** jrist has quit IRC | 23:14 | |
| *** andreykurilin has quit IRC | 23:18 | |
| *** yuanying has quit IRC | 23:20 | |
| *** yuanying has joined #openstack-ironic | 23:20 | |
| *** yuanying has quit IRC | 23:25 | |
| *** achanda has quit IRC | 23:27 | |
| *** achanda has joined #openstack-ironic | 23:27 | |
| *** achanda has quit IRC | 23:28 | |
| *** anderbubble has quit IRC | 23:32 | |
| *** penick has quit IRC | 23:35 | |
| *** yuanying has joined #openstack-ironic | 23:47 | |
| *** yuanying has quit IRC | 23:48 | |
| *** achanda has joined #openstack-ironic | 23:51 | |
| *** shakamunyi has quit IRC | 23:53 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!