Monday, 2014-05-19

« Sunday, 2014-05-18 Index Tuesday, 2014-05-20 »
*** lifeless has joined #openstack-ironic00:07
*** eghobo has quit IRC00:37
*** datajerk has joined #openstack-ironic00:41
*** nosnos has joined #openstack-ironic01:41
*** Haomeng has quit IRC01:49
*** datajerk has quit IRC01:51
*** lifeless has quit IRC02:16
*** matsuhashi has joined #openstack-ironic02:17
*** lifeless has joined #openstack-ironic02:24
*** killer_prince has quit IRC02:50
*** eghobo has joined #openstack-ironic03:19
*** lnxnut has joined #openstack-ironic03:30
*** coolsvap|afk is now known as coolsvap03:32
*** eghobo has quit IRC03:39
*** nosnos has quit IRC03:51
*** lnxnut has quit IRC03:57
*** john3213 has joined #openstack-ironic04:02
*** john3213 has left #openstack-ironic04:07
*** eghobo has joined #openstack-ironic04:20
*** nosnos has joined #openstack-ironic04:32
*** radsy has quit IRC05:19
*** Mikhail_D_ltp has joined #openstack-ironic05:36
*** Mikhail_D_ltp has quit IRC05:52
openstackgerritOpenStack Proposal Bot proposed a change to openstack/ironic: Imported Translations from Transifex  https://review.openstack.org/8850806:09
*** ramineni has joined #openstack-ironic06:18
*** adam_gone is now known as adam_g06:19
*** matsuhashi has quit IRC06:25
*** matsuhashi has joined #openstack-ironic06:28
*** eghobo has quit IRC06:40
*** eghobo has joined #openstack-ironic06:40
*** ifarkas has joined #openstack-ironic06:42
*** Mikhail_D_wk has joined #openstack-ironic06:51
Mikhail_D_wkGood morning Ironic! :)06:53
*** eguz has joined #openstack-ironic06:57
*** eghobo has quit IRC06:57
*** eguz has quit IRC06:59
dtantsurMorning Ironic07:02
*** geertj has joined #openstack-ironic07:23
*** foexle has joined #openstack-ironic07:35
*** jistr has joined #openstack-ironic08:01
*** killer_prince has joined #openstack-ironic08:15
*** athomas has joined #openstack-ironic08:19
*** _dguerri is now known as dguerri08:26
*** lucasagomes has joined #openstack-ironic08:32
*** martyntaylor has joined #openstack-ironic08:33
*** derekh_ has joined #openstack-ironic08:37
*** dguerri is now known as _dguerri08:52
*** _dguerri is now known as dguerri09:01
*** ndipanov has joined #openstack-ironic09:16
openstackgerritLucas Alvares Gomes proposed a change to openstack/python-ironicclient: Add pagination support to {node, port, chassis}-list  https://review.openstack.org/9158509:22
*** lsmola has joined #openstack-ironic09:44
*** matsuhashi has quit IRC09:50
*** geertj has quit IRC09:57
*** geertj has joined #openstack-ironic10:07
*** martyntaylor has left #openstack-ironic10:07
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Remove 'node' parameter from the Console and Rescue interfaces  https://review.openstack.org/9109710:08
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Remove 'node' parameter from the Power interface  https://review.openstack.org/9109310:09
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Remove 'node' parameter from the Deploy interface  https://review.openstack.org/9106210:09
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Remove 'node' parameter from the validate() methods  https://review.openstack.org/9110410:09
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Remove 'node' parameter from the Console and Rescue interfaces  https://review.openstack.org/9109710:15
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Remove 'node' parameter from the validate() methods  https://review.openstack.org/9110410:15
*** coolsvap is now known as coolsvap|afk10:16
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Pass kwargs to ClientWrapper's call() method  https://review.openstack.org/9238410:29
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: list_instances() to return a list of instances names  https://review.openstack.org/9238510:29
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Add the remaining unittests to the ClientWrapper class  https://review.openstack.org/9241610:35
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Run keepalive in a dedicated thread  https://review.openstack.org/9308310:39
*** ramineni has quit IRC11:01
*** dguerri is now known as _dguerri11:12
*** lnxnut has joined #openstack-ironic11:13
*** overlayer has joined #openstack-ironic11:16
*** nosnos has quit IRC11:21
*** lnxnut has quit IRC11:24
NobodyCamGood morning Iroinc11:45
agordeevNobodyCam: good morning11:46
agordeevfor everyone else good morning too!11:46
lucasagomesNobodyCam, hey buddy! worrying about u cause u didn't appear in the summit on the last days11:47
lucasagomesNobodyCam, everything is ok?11:47
lucasagomesagordeev, morining11:47
Shrewsmorning all11:49
lucasagomesmorning Shrews11:49
NobodyCammorning agordeev lucasagomes and Shrews11:50
NobodyCamlucasagomes: I had to leave the summot to deal with family issues... :(11:50
lucasagomesNobodyCam, :( I c, I hope everything is fine by now11:51
NobodyCammy dad had a stroke11:51
lucasagomesNobodyCam, oh, sorry to hear that11:51
NobodyCamI am about two hours out side NEbreaska11:51
agordeevlucasagomes, Shrews morning :)11:51
NobodyCamwill be htere today11:51
lucasagomesNobodyCam, :)11:52
lucasagomesgood good11:52
NobodyCambrb walkies and coffee11:53
lucasagomesack, I'm going for lunch brb as well11:58
*** lucasagomes is now known as lucas-hungry11:58
*** pelix has joined #openstack-ironic11:59
geertjQuestion: is there a place to store driver specific information that is not node specific?12:00
NobodyCamgeertj: such as?12:04
geertjNobodyCam: I'm trying to do a "virtual bare metal" driver for Ravello. I need a global API username, password, and some more metadata.12:04
Shrewsgeertj: config options?12:07
openstackgerritAleksandr Gordeev proposed a change to openstack/ironic-python-agent: Add hardware_utils  https://review.openstack.org/9239912:08
NobodyCamgeertj: I don't think we have that today.12:08
openstackgerritAleksandr Gordeev proposed a change to openstack/ironic-python-agent: Introduce MainlineHardwareManager  https://review.openstack.org/9284712:08
NobodyCamya I could see a config option12:08
geertjNobodyCam/Shrews: I think config options are the right solution.12:09
geertjI see there's options for "ssh" and "seamicro" too.12:09
Shrewsyup12:09
NobodyCam:)12:11
NobodyCamI've also envisioned chassis holding defaults for nodes. but not yet12:12
NobodyCam:0p12:12
*** Mikhail_D_wk has quit IRC12:12
*** Mikhail_D_wk has joined #openstack-ironic12:13
geertjOptions in a chassis would actually be useful here. Ravello has the notion of an application which would map very well to a chassis.12:13
geertjBut I can look into that later. First want to get something up and running :)12:13
NobodyCamgeertj: yep :)12:13
geertjRegarding the options: it looks like etc/ironic/ironic.conf.sample is autogenerated? Any hints?12:13
NobodyCam??? add the option and regrn the sample12:14
NobodyCamadd the option to the code and re gen the sample congig12:14
NobodyCamgah ... /me needs coffee12:15
geertjThis is all new to me... Where do I add define an option (sorry for asking noob questions...)12:15
NobodyCamgeertj: taake a look at something like the ssh driver. https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/ssh.py#L41-L4812:16
geertjOK that is simple.. (*blush*)12:17
NobodyCam:)12:18
Shrewsugh, just tried the new gerrit change view. *vomit*12:18
*** jdob has joined #openstack-ironic12:20
*** krtaylor has joined #openstack-ironic12:22
*** early has quit IRC12:29
*** dhellmann is now known as dhellmann_12:31
*** early has joined #openstack-ironic12:32
*** lucas-hungry is now known as lucasagomes13:00
*** dlfryar has joined #openstack-ironic13:00
*** jbjohnso has joined #openstack-ironic13:10
anteayahey ironic, a third party meeting is now a thing: https://wiki.openstack.org/wiki/Meetings/ThirdParty13:12
anteayaso if you or anyone you love should be there13:13
anteayado share the info13:13
NobodyCamTy anteaya13:18
NobodyCamand good mornignn13:18
NobodyCammorning even13:18
*** dlfryar has quit IRC13:18
anteayamorning13:22
anteayasorry I didn't get to meet the kids13:22
anteayaI hope you got things smoothed out?13:22
NobodyCamty :( ya did quite go as I planed.. :(13:23
NobodyCamwill know more later tonight / today. just about two hours out side my dads town now.13:23
NobodyCamworking on some tests :-p13:24
anteayak13:24
anteayaI hope things go well13:24
anteayagood luck13:24
anteaya:D13:24
NobodyCamTY.. :/13:24
anteayayeah :/13:24
*** killer_prince has quit IRC13:34
*** romcheg has joined #openstack-ironic13:46
*** matty_dubs|gone is now known as matty_dubs13:48
NobodyCamlucasagomes: can I ask you a mock question13:50
lucasagomesNobodyCam, sure, I'm not expert in mock tho13:51
openstackgerritAndrey Kurilin proposed a change to openstack/ironic: Renaming ironicclient exceptions in nova driver  https://review.openstack.org/9004013:52
NobodyCamlucasagomes: http://paste.openstack.org/show/LdvqET2tPvG7rZdbwSNp/13:53
lucasagomesNobodyCam, seems that there's a "seek=-35" that is being passed to the dd command13:54
lucasagomeswhich is not part of the expected command list13:54
*** romcheg has quit IRC13:54
NobodyCamam I setting up that mock correctly?13:54
NobodyCamhumm13:55
NobodyCamhttp://paste.openstack.org/show/p3E4aoe2YVn4oGiXTzBU/13:55
lucasagomesNobodyCam, ahh, so that function runs 2 dd's?13:55
lucasagomesso there's a way to test it, lemme find an e.g for u13:56
NobodyCamthats the code it should be testing13:56
jrollmorning ironic13:56
*** annegent^ has joined #openstack-ironic13:57
NobodyCammorning jroll13:57
lucasagomesNobodyCam, https://github.com/openstack/ironic/blob/master/ironic/tests/drivers/test_ipmitool.py#L247-L25313:57
jrollNobodyCam: the tldr is you make a list of calls, like calls = [mock.call(arg1, arg2=arg2), mock.call(...)], then use mock_func.assert_has_calls(calls)13:58
jrollor that works too13:58
lucasagomesjroll, yeah, we use assertEqual to check for the calls13:58
jrollah, ok13:58
lucasagomesjroll, but assert_has_calls seems grand as well13:58
jrollyeah, does the same thing13:59
*** jrist has joined #openstack-ironic13:59
NobodyCamahh TY14:00
jrollNobodyCam: one more thing that is nice to know, you can use mock.ANY as an expected argument to allow any value for that arg (like for a random number or something)14:00
*** killer_prince has joined #openstack-ironic14:02
anteayaso when devananda is available, do you actually want a separate gerrit group for your specs repo? ironic-specs-core? https://review.openstack.org/#/c/94113/2/modules/openstack_project/files/gerrit/acls/openstack/ironic-specs.config14:03
anteayaI just want to ensure this is what you want to have happen14:03
matty_dubsMorning all!14:04
NobodyCammorning matty_dubs14:04
*** rwsu has joined #openstack-ironic14:06
*** linggao has joined #openstack-ironic14:06
NobodyCambrb... quick walkies14:07
geertjDo Ironic drivers run in their own thread? Is it OK to call blocking network calls?14:08
devanandamorning, all14:10
lucasagomesdevananda, morning14:11
devanandageertj: no. ironic-conductor is currently using greenthreads, not real threads14:11
geertjdevananda: ouch...14:12
devanandaanteaya: it seems reasonable to me to allow separate groups to review code vs. design14:12
devanandageertj: that said, eventlet handles most network io in a non-blocking way14:13
geertjdevananda: I need to call a restful API service. I've got an wrapper that is using Python's httplib.14:13
devanandageertj: that should be fine14:13
geertjdevananda: OK.. so eventlet is patching the stdlib?14:14
jrollmorning devananda14:14
*** lnxnut has joined #openstack-ironic14:14
linggaomorning, ironic folks.14:16
anteayadevananda: great, as long as this is intentional, I am good with that14:16
devanandageertj: http://eventlet.net/doc/patching.html#monkeypatching-the-standard-library14:16
anteayadevananda: chmouel did find an oops with the test file on that patch though, so I will wait for a new patchset14:17
geertjdevananda: kewl14:17
devanandaanteaya: ack, looking14:17
anteayaand hello everybody14:17
anteayahey matty_dubs14:17
anteayaI just love saying your username14:17
matty_dubsHowdy anteaya14:17
matty_dubsHaha14:17
anteaya:D14:18
lucasagomesmorning linggao matty_dubs14:19
lucasagomesanteaya, and geertj as well :)14:19
matty_dubsHeya lucasagomes14:19
anteayahey lucasagomes14:19
geertjmorning lucasgomes :)14:19
linggaomorning devananda, lucasagomes, matty_dubs, NobodyCam14:20
matty_dubsHi linggao!14:20
matty_dubsSo, as a Rubyist reluctantly working in Python, I got a big chuckle out of seeing this book at one of the bookstores in the Atlanta airport: http://www.amazon.com/The-Spirit-Python-Exposing-Squeeze/dp/162136220514:21
NobodyCammornig linggao devananda14:21
linggaoNobodyCam, are you still on the road?14:22
lucasagomesmatty_dubs, lol hah14:23
lucasagomesmatty_dubs, did you buy it?14:24
jrollmatty_dubs: lolol14:24
linggaomatty_dubs, lol.14:25
NobodyCamlinggao: yes I am in Kansas City .. about two hours out from where my dad lives14:26
linggaoNobodyCam, cool! enjoy!14:26
*** lazy_prince3 has joined #openstack-ironic14:30
*** lazy_prince3 is now known as lazy_prince14:31
devanandahmmm well, we're skipping our meetnig today, BUT anteaya is hosting the inaugural third-party-CI meeting today14:44
devanandaso if anyone is around and can attend, that'd be great -- details here -- https://wiki.openstack.org/wiki/Meetings/ThirdParty14:44
anteayathanks devananda14:45
anteayaall welcome14:45
*** lnxnut is now known as hipster14:45
openstackgerritChris Krelle proposed a change to openstack/ironic: Workaround for conflicting volume labels  https://review.openstack.org/9313314:49
NobodyCamany one want to take a look at ^^^14:49
NobodyCamlucasagomes: can you look at the test and let me know how you think they look14:49
NobodyCambbt...brb14:51
*** jgrimm has joined #openstack-ironic14:51
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Speed up Nova Ironic driver tests  https://review.openstack.org/9420214:51
lucasagomesNobodyCam, sure, will do14:51
NobodyCam:) TY14:52
*** blamar has joined #openstack-ironic15:07
NobodyCambrb15:07
*** hipster has quit IRC15:08
pelixtrying to get ironic running via devstack by following http://docs.openstack.org/developer/ironic/dev/dev-quickstart.html#deploying-ironic-with-devstack15:22
pelixrunning into a brick wall around getting the baremetal nodes created to come up15:22
pelixseems no IP address is available for them from DHCP provided by neutron/dnsmasq when booting15:22
pelixany suggestions on where to look to see what's going wrong?15:23
NobodyCampelix: is there a firewall running?15:25
pelixNobodyCam: no, the dnsmasq program sees the request but logs "dnsmasq-dhcp[16004]: DHCPDISCOVER(tapde587ec3-65) 52:54:00:01:f9:25 no address available" to syslog and doesn't respond15:28
pelixadding an IP address manually to the hosts file and issuing a SIGHUP to the process 16004 and resetting the domain allows the ipaddress to be assigned, but then it reports no file found (I'm guessing as in no pxelinux.0 file found)15:29
pelixso it seems like something earlier hasn't done the expected set up both for neutron to provide the address and for the tftpboot area to be properly available15:29
*** mkerrin has joined #openstack-ironic15:30
pelixbtw, although ufw reports as inactive, there is some output from 'iptables -L'15:31
pelixbut since it isn't blocking the DHCP request or response I'm assuming whatever being done regarding firewall conf, is working as expected (tftp port is open as well).15:32
NobodyCamhumm15:34
lucasagomesNobodyCam, reviewed15:34
pelixNobodyCam: has anything changed that renders the conf listed in the docs as obsolete?15:35
NobodyCamlucasagomes: TY15:37
*** igor_ has joined #openstack-ironic15:37
lucasagomesNobodyCam, np15:37
lucasagomesleft some inline comments15:38
NobodyCampelix: not that I know of15:38
NobodyCamyep will address now15:38
*** dlfryar has joined #openstack-ironic15:39
devanandain case folks didn't see the mail yet, i've started working on our specs repo15:42
devanandafor the moment, it's on my github (pending infra adding it to openstack)15:42
devanandahttps://github.com/devananda/ironic-specs/blob/master/specs/template.rst15:42
*** Mikhail_D_ltp has joined #openstack-ironic15:44
*** matty_dubs is now known as matty_dubs|lunch15:46
lucasagomesdevananda, the template looks good... one thing15:47
lucasagomesdevananda, in the performance impact section you say that periodic tasks runs in a single greenthread etc... this is true for now, but it's going to be fixed no? I remember you mentioning that there is some ongoing work on nova to fix that am I correct?15:48
lucasagomes(just thinking if it worth to be mentioned in the template if it's going to be fixed soonish)15:49
devanandalucasagomes: we can always change the template when it gets fixed15:50
lucasagomesdevananda, sure, ack... good stuff!15:51
lucasagomesthe template lgtm15:51
*** hemnafk is now known as hemna15:54
devanandalucasagomes: also https://github.com/devananda/ironic-specs/blob/master/specs/juno/fully-async-api.rst15:57
*** eghobo has joined #openstack-ironic15:58
*** datajerk has joined #openstack-ironic15:58
lucasagomesdevananda, nice will read in a sec15:58
devanandajroll: ping16:01
lucasagomesdevananda, I was thinking about creating the concepting of task for those async calls... e.g16:02
lucasagomes PUT (or POST even) {'enabled': 'true'}  /v1/nodes/<uuid>states/console, HTTP CODE: 202, LOCATION /v1/nodes/<uuid>states/console/a66h1216:03
lucasagomesthis way, the task is traceable via /v1/nodes/<uuid>states/console/a66h1216:03
lucasagomesso use can issue a GET on that URI to get the status of it16:03
lucasagomes(didn't think much about it) just an initial idea16:04
devanandalucasagomes: that leads to constant polling by clients, not very scalable16:06
lucasagomesdevananda, yeah :/ as-is today's16:06
lucasagomesdevananda, maybe a callback mechanism would be better16:06
devanandalucasagomes: and managing those resources end-of-life as well16:06
devanandawe talked in the nova track about cross service callback mechanisms16:07
devanandanova<->neutron already does this a little bit16:07
devanandaat some point, we'll want to be able to call back to nova to say "this request is done" rather than have nova continuously poll status16:07
lucasagomesdevananda, neat... yeah, I was just thinking because that would offer some flexibility when managing specific tasks, like being able to issue DELETE to abort a giving task etc16:07
lucasagomesyeah the callback seems pretty good and scalable16:08
devanandait's good for inter-service communication. not so much with HTTP clients16:08
openstackgerritChris Krelle proposed a change to openstack/ironic: Workaround for conflicting volume labels  https://review.openstack.org/9313316:09
devanandahow would a CLI user receive a callback when they are, say, behind a NAT firewall?16:09
NobodyCamlucasagomes: ^^^ :)16:09
devanandalucasagomes: here's another idea16:10
lucasagomesNobodyCam, will take a look in a bit, finishing something up quickly16:10
devananda* allow API service to lock nodes, as part of precondition checking16:11
devananda* API updates "target_*" states when requested to do something (eg, start a deploy)16:11
devananda* API then "cast"s a notice to the conductor, and responds with 202 to user16:11
devanandathere's no synchronous RPC call at all, but the user knows the action can start16:11
NobodyCam:)16:12
devanandaif API is unable to lock the node w/in some timeout (less than HTTP timeout), user gets a timeout error16:12
devanandaand if API gets a lock but the state check fails, user gets a notice as to why their request couldn't be performed16:12
devanandathis needs no change in the REST API usage or error codes, I think...16:13
devanandaactually, one positive change - clients won't need to retry for NodeLocked errors any more16:15
devanandarussell_h: ping16:15
lucasagomesdevananda, should the API service do the locking? it would require moving the task manager to a common place16:15
russell_hdevananda: hey16:15
devanandalucasagomes: yes16:15
lucasagomesI mean, it could be done via RPC as well16:15
devanandarussell_h: hey there. see my mail to the list re: specs repo?16:15
NobodyCamok last of the road trips (at least for a while). Kansas city to Omaha16:15
russell_hso locks get passed via RPC?16:15
devanandalucasagomes: doign that over RPC leads t othe problem we have today16:16
devanandarussell_h: I was pinging you re: specs, not this conversation -- we're multitasking :)16:16
russell_hheh, yeah, sorry I was just catching up on this conversation16:16
*** jistr has quit IRC16:16
russell_hI was experimenting with refactoring locking yesterday16:16
russell_hso I'm interested16:16
russell_hanyway16:16
devanandarussell_h: pending infra adding it to openstack/irinic-specs, I've got draft up here https://github.com/devananda/ironic-specs/blob/master/specs/template.rst16:17
russell_hre: specs, didn't read the email yet16:17
lucasagomesyeah, hmm def the idea worth a go16:17
russell_hbut I love the idea if its what I think it is16:17
devanandarussell_h:  i'm soliciting input on whether the template covers all the things. And suggesting that ya'll start spec'ing the things that need to be spec'd16:17
*** hipster has joined #openstack-ironic16:18
devanandarussell_h: in particular, the refactoring for instance_info and pulling out common methods for image and dhcp manipulation16:18
devanandarussell_h: as those will need to land before the IPA driver anyway16:18
russell_hseems reasonable16:18
russell_halright, I'll take a look at the template in an hour or so16:19
devanandarussell_h: thanks much16:19
russell_hand put the team on getting specs up over the next few days16:19
*** geertj has quit IRC16:21
lucasagomesbtw, the tests we currently have on devstack it doesn't issue any nova boot right?16:22
lucasagomesdevananda, ^16:22
lucasagomesNobodyCam, ^, would the check-tripleo-ironic-undercloud-precise do that?16:23
devanandalucasagomes: ironic-virtual test does a full deploy and SSH into the instane16:26
devanandawe should make that voting soon, IMO16:26
*** newell_ has joined #openstack-ironic16:26
lucasagomesdevananda, hmm I thought it did as well but idk now, because I was reviewing the refactoring patch (https://review.openstack.org/90233)16:27
lucasagomesand it's broken because it's not generating the pxe config files (the function is not being called)16:28
lucasagomesbut the tests passed16:28
russell_hdevananda: lucasagomes: I was experimenting some with making locking pluggable yesterday: https://gist.github.com/russellhaering/5035399c523cacbca60416:30
devanandalucasagomes: no they didn't -- look at http://logs.openstack.org/33/90233/34/check/check-tempest-dsvm-virtual-ironic/fe0a14016:30
devanandalucasagomes: that's exactly why it should be made voting :)16:30
lucasagomesdevananda, ah damn! I was looking at the tripleo ones16:31
lucasagomesdevananda, +1 for making it voting! _much_ needed16:31
lucasagomesrussell_h, will take a look16:32
russell_hlucasagomes: it may not be quite what we want if we're planning to rework how locking works, but my goal was basically to make it to where I could plug in zookeeper for locking16:33
russell_hand eventually for conductor registration as well16:33
lucasagomesrussell_h, I see, it might worth to spend some time on the meeting (or even here on IRC) to think about hw the locks and new api model should work and build the specs for it16:35
lucasagomeswould be good to refactor that out asap like j1 or start of j2 to get it all in trunk16:35
russell_hyes16:35
lucasagomeson IRC I mean this channel16:36
russell_hyeah, I figured16:36
russell_hthats actually better for me16:36
russell_hI'm triple-booked during the weekly meeting16:36
russell_hevery week16:36
russell_hits great16:36
lucasagomesoh I c, it works for me here as well16:36
JoshNangi thought there wasn't a meeting this week anyway?16:36
lucasagomesJoshNang, there's not16:36
lucasagomesbut we could add it to the agenda for the next if that's the case16:37
JoshNang+116:37
lucasagomesand put some initial thoughts somewhere where people can read before attending to it16:37
russell_hI'd rather put more rigor into it before the meeting16:38
russell_hso we're basically proposing a detailed plan16:38
russell_hbut, I just don't like meetings16:38
russell_hso either way :)16:38
lucasagomesheh, russell_h it's odd for me to hear a manager saying he/she doesn't like meetings :)16:39
lucasagomesI agree16:39
lucasagomesalthough the ironic one, 1 hour per week is being beneficial16:39
russell_hyeah, I usually end up just reading the backlog, but it seems decent16:40
devanandarussell_h: we can move the meeting - you're not the only one for whom its not a good time.16:42
devanandarussell_h: though the openstack meeting calendar is preciously full already...16:44
jrolldevananda: pong16:44
*** foexle has quit IRC16:45
devanandajroll: hi! was just pinging you re: the specs repo, same as I pinged russell_h. see ML or scrollback16:45
devanandarussell_h: re: pluggable lock model -- yes please16:45
*** igor_ has quit IRC16:45
devanandarussell_h: as far as plan vs. irc discussion, that's what the specs repo is for :)16:45
jrolldevananda: reading scrollback now :)16:45
*** igor_ has joined #openstack-ironic16:45
devanandathen we discuss on the spec review rather than during the already-crammed weekly meeting16:46
russell_h+1million16:46
lucasagomesthere's only one openstack-meeting channel?16:46
russell_hI basically ran into the question of how to implement "blocking for a lock", given that we must do that inline in the API call16:46
devanandathere's two now16:46
lucasagomeswouldn't be better to create more maybe? so time would be more flexible for projects16:46
lucasagomesI see16:46
devanandamaybe three, heh16:46
jrolldevananda: your email said 'I will focus on creating and landing the specs for items essential for graduation first...' - I'm assuming you're not creating *all* of the specs, just the ones you're responsible for or something. like, I'll create the agent specs etc16:46
lucasagomesI see, yeah that would be useful16:46
devanandayep, there's 3, heh16:47
devanandajroll: agent isn't req'd for graduation16:47
devanandajroll: but i'm happy if others create the specs for the things that are req'd if they want to16:47
jrolldevananda: oh, wow, I totally read that as 'critical for juno'16:48
* jroll chugs coffee16:48
russell_hI see two options: a) block only for a short period of time, then error out, or b) add a task-tracking API so we can accept API calls, then actually execute the associated task when the node frees up16:48
devanandarussell_h: a++16:48
*** harlowja has joined #openstack-ironic16:48
devanandab--16:48
russell_hdevananda: do we actually need to do the locking in the API? Is it OK to still do it in the conductor, just with a timeout known to be much shorter than the API timeout?16:49
devanandarussell_h: get lock; validate pre-state; set target-state in db; unlock; cast to conductor; return 20216:49
devanandawell16:50
devanandai'm still being too pessimistic with locking for that to scale well16:50
russell_hif the validation happens in the API, every API will need to load every driver16:50
*** igor_ has quit IRC16:50
russell_hin real life thats probably fine16:50
russell_hbut a little different from today's model16:50
devanandahm. that's not the validation i'm referring to16:51
devanandaso let's think about how this could be done with no lock in the API16:51
russell_hoh, you just mean the stuff thats already in the API16:51
*** matty_dubs|lunch is now known as matty_dubs16:51
*** martyntaylor has joined #openstack-ironic16:51
russell_hcommon validation of node state and so on16:51
lucasagomesdevananda, between the unlock and the cast some periodic task could then take the lock from the conductor16:51
devanandaright16:51
lucasagomesand then we would have a dirty target_state and no conductor working on it16:52
russell_hso if we moved that out of the API, and into the conductor16:52
russell_hto happen _after_ the lock is achieved16:52
devanandalucasagomes: yep. so periodic tasks would need to skip that node during that interval16:52
*** derekh_ has quit IRC16:52
lucasagomesdevananda, would be better to keep the lock?16:52
devanandawe need an intent-lock anyhow16:52
jrollI want to point out that validation on node state sometimes needs to happen in the driver16:52
devanandaalso for the external scheduler bits we discussed16:52
jrolle.g. power_state validation for do_node_deploy16:52
lucasagomesdevananda, and use a call with a greenthread to guaratee that the conductor got the task?16:52
lucasagomesyeah that's an intent lock16:53
lucasagomesjroll, I think that the validation devananda is referering to is just checking the current db value, not issuing any driver-specific commands16:54
russell_hdevananda: you could also _not_ unlock in the API, and actually "hand off" the lock to the conductor16:54
russell_hthere are... risks16:55
lucasagomesrussell_h, yeah, that's the intent lock16:55
lucasagomesI agree with it16:55
devanandaextrapolating all this out, we're talking about a distributed state machine. FSMs are a blight on existence. DFSMs are unspeakable.16:55
russell_hheh16:55
jrolllucasagomes: you might be right, but that's an example of something stored in the db that validation is performed on16:55
devanandado we actually need one, or just /think/ we do?16:55
russell_hwell16:55
russell_hthere is another option16:56
russell_ha conductor could lock every node it thinks it owns, just up front16:56
russell_hthen schedule work against that node in-process16:56
russell_hif the conductor manages all state, its not distributed16:56
russell_hexcepting between the conductor and the db16:56
russell_h(well, and the BMC and...)16:57
devanandaapi service then casts "a request to do something"16:57
devanandaand the conductor may choose to ignore it if it's unservicable (like "deploy node X" when node X is already deployed)16:57
russell_hsomething like that, yes16:58
devanandait resolves all the locking and race conditions cleanly. however the user experience will be more like "hey, where'd my request go"16:58
russell_hso, you don't have to rely on cast16:58
devanandaunless there's some state checking in the API16:58
devanandaeven if that's async, non-locked, and possibly stale16:59
devanandait's probably good enough16:59
russell_hyou can have the conductor either respond "yes, will do", or "no, not possible"16:59
*** martyntaylor has left #openstack-ironic16:59
russell_hbasically what happens today16:59
devanandarussell_h: which is what i'm proposing to avoid16:59
russell_hexcept that everything is in-memory in a single conductor16:59
russell_hright17:00
devanandaanyhow17:00
devanandai'd rather we discuss all this in a review on the specs repo, once it's created17:00
*** killer_prince has quit IRC17:00
russell_hagreed17:00
devanandaso there's a trail for future reference (aside from irc logs)17:00
lucasagomes+117:01
jrolldevananda: can you ping me when the specs repo exists, and I'll get those agent specs going?17:10
lucasagomesalright, it's dinner time for me17:13
lucasagomeshave a good night everyone!17:14
matty_dubsAdios, lucasagomes!17:14
*** lucasagomes is now known as lucas-hungry17:14
jroll\o17:15
*** igor_ has joined #openstack-ironic17:16
*** igor_ has quit IRC17:21
*** athomas has quit IRC17:35
*** killer_p- has joined #openstack-ironic17:39
*** killer_p- is now known as killer_prince17:40
*** lazy_prince has quit IRC17:40
*** geertj has joined #openstack-ironic17:41
devanandajroll: you can start from what i've got here: https://github.com/devananda/ironic-specs/blob/master/specs/template.rst17:46
openstackgerritJosh Gachnang proposed a change to openstack/ironic: Adding a reference driver for the agent  https://review.openstack.org/8479517:46
devanandajroll: i'll notify the ML when the repo gets moved to openstack/17:46
jrolldevananda: cool, that was my plan. thanks17:46
*** harlowja has quit IRC17:50
*** harlowja has joined #openstack-ironic17:50
*** harlowja has quit IRC18:00
*** harlowja has joined #openstack-ironic18:00
*** datajerk has quit IRC18:00
*** geertj has quit IRC18:08
*** datajerk has joined #openstack-ironic18:12
*** hipster has quit IRC18:15
*** igor_ has joined #openstack-ironic18:16
*** datajerk has quit IRC18:17
*** igor_ has quit IRC18:21
*** eghobo has quit IRC18:29
*** datajerk has joined #openstack-ironic18:34
*** pelix has quit IRC18:34
*** tatyana has joined #openstack-ironic18:37
openstackgerritDavid Shrewsbury proposed a change to openstack/ironic: Update dev quickstart guide for ephemeral testing  https://review.openstack.org/9210518:39
*** datajerk has quit IRC18:41
*** lazy_prince has joined #openstack-ironic18:42
*** hipster has joined #openstack-ironic18:47
*** hipster has quit IRC18:51
*** eghobo has joined #openstack-ironic18:59
*** geertj has joined #openstack-ironic19:00
*** igor_ has joined #openstack-ironic19:17
*** hipster has joined #openstack-ironic19:18
*** igor_ has quit IRC19:22
*** hipster has quit IRC19:23
*** dlfryar has quit IRC19:29
matty_dubsSo, dumb question time. I see 'dsvm' in a lot of the tests. What does that stand for?19:35
matty_dubsGoogle is completely unhelpful and keeps showing the DSM 5 when I try to search, which isn't helping.19:35
*** eguz has joined #openstack-ironic19:37
*** eghobo has quit IRC19:41
*** lazy_prince has quit IRC19:46
*** hipster has joined #openstack-ironic19:47
Shrewsdevstack vm?19:49
Shrewswhere do you see it?19:49
*** epim has joined #openstack-ironic19:49
*** epim is now known as Penick19:50
matty_dubsJust in test names, like check-tempest-dsvm-ironic19:51
matty_dubs'devstack vm' makes perfect sense19:51
matty_dubsBut somehow never occurred to me19:51
*** hipster has quit IRC19:52
*** greghaynes has quit IRC19:53
Shrewswe don't have nearly enough acronyms19:53
*** christopheraedo has quit IRC19:54
*** zul has quit IRC19:54
*** ewindisch has quit IRC19:54
*** soren has quit IRC19:54
*** jrist has quit IRC19:55
*** tteggel has quit IRC19:55
*** tteggel has joined #openstack-ironic19:55
*** christopheraedo has joined #openstack-ironic19:55
*** soren has joined #openstack-ironic19:56
*** geertj has quit IRC19:56
*** zul has joined #openstack-ironic19:58
*** harlowja has quit IRC20:01
*** tatyana has quit IRC20:02
*** harlowja has joined #openstack-ironic20:02
*** foexle has joined #openstack-ironic20:03
*** jrist has joined #openstack-ironic20:06
*** tatyana has joined #openstack-ironic20:07
*** greghaynes has joined #openstack-ironic20:09
*** ewindisch has joined #openstack-ironic20:16
*** igor_ has joined #openstack-ironic20:18
*** tatyana has quit IRC20:21
*** dlfryar has joined #openstack-ironic20:22
*** igor_ has quit IRC20:23
*** foexle has quit IRC20:29
*** Mikhail_D_ltp has quit IRC20:33
*** lazy_prince has joined #openstack-ironic20:34
*** Hefeweizen has joined #openstack-ironic20:35
*** hipster has joined #openstack-ironic20:47
*** linggao has quit IRC20:48
*** hipster has quit IRC20:52
Shrewsumm, the check-tripleo-ironic-* tests are coming back NOT_REGISTERED. is that new or have i been totally unobservant?20:54
*** jdob has quit IRC20:59
*** gmatefi has joined #openstack-ironic20:59
*** jbjohnso has quit IRC21:08
devanandaShrews: i think infra changed something in how things report recently21:09
devanandaalso, it looks like dsvm-virtual-ironic has may be failing all the time now??21:10
adam_ghmm21:11
adam_ghttp://no-carrier.net/~adam/openstack/ironic_gate_status.html21:11
devanandaadam_g: hm, ok. not all the time - just the ones I randomly looked at21:13
* devananda bookmarks that21:13
devanandaadam_g: any thoughts on how to determine what % of those are false negatives?21:14
adam_gdevananda, hmm. looks like the tempest run is passing, but some logs have non-whitelisted errors in them21:16
adam_gie https://jenkins05.openstack.org/job/check-tempest-dsvm-virtual-ironic-nv/369/console21:16
devanandaI see. both of those are expected.21:18
devananda*should be21:18
devanandaand should be present in every run21:18
adam_gdevananda, the mac addr errors?21:18
*** igor_ has joined #openstack-ironic21:19
devanandayep21:21
devanandathose come from a tempest test21:21
devanandathough, i had a patch up to suppress logging them21:22
*** igor_ has quit IRC21:23
*** matty_dubs is now known as matty_dubs|gone21:27
openstackgerritDevananda van der Veen proposed a change to openstack/ironic: Implement the Conductor side of ring rebalancing  https://review.openstack.org/7919421:30
*** Penick has quit IRC21:31
devanandamrda-summit: found some old code I was working on for the hash ring -- https://review.openstack.org/#/c/79194/ -- this may give you a starting point. or may be totally wrong and should be re-buried21:31
*** harlowja has quit IRC21:32
openstackgerritDevananda van der Veen proposed a change to openstack/ironic: Don't update neutron VIF if MAC is duplicated  https://review.openstack.org/9043621:33
devanandaadam_g: yea, this patch should hide the MACAlreadyExists error -- https://review.openstack.org/#/c/90436/ but the patch may need a bit more work21:33
*** harlowja has joined #openstack-ironic21:33
adam_gdevananda, cool. i dont see how all runs are not failing on that error, since it should be logged everytime the negative test runs?21:34
devanandaright21:35
devanandai dont understand that yet either21:35
adam_gAFAICS the ironic logs errors are failing the test, not any of the others that appear not to be whitelisted21:35
*** hipster has joined #openstack-ironic21:47
*** lazy_prince has quit IRC21:48
NobodyCamHello from Greenwood Ne21:50
devanandaNobodyCam: o/21:51
*** hipster has quit IRC21:52
NobodyCam\o devananda21:52
NobodyCamdevananda:(and all others) have a minute to look at https://review.openstack.org/#/c/9313321:54
*** igor_ has joined #openstack-ironic21:58
*** igor_ has quit IRC22:03
*** foexle has joined #openstack-ironic22:10
NobodyCamty devananda will get a new patch up in a bit to address the comments/ conserns :)22:11
*** radsy has joined #openstack-ironic22:31
*** jgrimm has quit IRC22:38
*** gmatefi has quit IRC22:38
*** hipster has joined #openstack-ironic22:42
*** hipster_ has joined #openstack-ironic22:44
*** hipster has quit IRC22:47
*** hipster has joined #openstack-ironic22:47
*** hipster_ has quit IRC22:49
openstackgerritChris Krelle proposed a change to openstack/ironic: Workaround for conflicting volume labels  https://review.openstack.org/9313322:49
*** r-daneel has joined #openstack-ironic22:50
NobodyCamdevananda: see if that addresses the conserns... maybe to hackish22:50
*** hipster has quit IRC22:52
*** overlayer has quit IRC22:57
*** igor_ has joined #openstack-ironic22:59
*** r-daneel has quit IRC23:03
*** igor_ has quit IRC23:03
*** r-daneel has joined #openstack-ironic23:04
*** r-daneel has quit IRC23:07
*** datajerk has joined #openstack-ironic23:17
*** datajerk has quit IRC23:22
*** datajerk has joined #openstack-ironic23:23
*** lucas-hungry has quit IRC23:24
*** matsuhashi has joined #openstack-ironic23:27
*** datajerk has quit IRC23:28
*** romcheg has joined #openstack-ironic23:38
*** romcheg1 has joined #openstack-ironic23:40
*** romcheg has quit IRC23:42
jrolldevananda: this landed a while ago, can we kill the blueprint somehow? https://blueprints.launchpad.net/ironic/+spec/add-node-instance-info23:44
jrolldevananda: also, do you want a spec for moving things driver_info -> instance_info? instinct says yes23:44
*** dlfryar has quit IRC23:45
*** hipster has joined #openstack-ironic23:47
*** hemna is now known as hemna_23:49
*** hipster has quit IRC23:52
« Sunday, 2014-05-18 Index Tuesday, 2014-05-20 »

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!