Wednesday, 2022-07-13

jandersgood morning Ironic o/00:05
TheJuliao/00:30
opendevreviewMerged openstack/metalsmith stable/yoga: Add linters job  https://review.opendev.org/c/openstack/metalsmith/+/84952303:10
opendevreviewMerged openstack/metalsmith stable/wallaby: Add linters job  https://review.opendev.org/c/openstack/metalsmith/+/84958303:19
opendevreviewMerged openstack/metalsmith stable/xena: Add linters job  https://review.opendev.org/c/openstack/metalsmith/+/84952403:19
opendevreviewAija Jauntēva proposed openstack/ironic master: Fix markup typo in Redfish driver docs  https://review.opendev.org/c/openstack/ironic/+/84961706:56
kaloyanTheJulia: Yes, I was refering to ipxe configuration sym-links having names in /httpboot. Maybe I don't understand well enough how iPXE works but it was a bit misleading having a message saying: "Attempting to boot from MAC X", where X refers to a NIC that's not PXE-capable (there's no DHCP in that VLAN).07:09
kaloyanTheJulia: It turned out it was caused by the iSCSI SAN having specific requirements when using L3 networks. Placing its portals in the same VLAN and subnet resolves the problem.07:13
kaloyanTheJulia: It "resolves" the problem, because iPXE can now talk directly via the configure interface, regardless of the message output07:16
kaloyanTheJulia: My point is that having iPXE configuration in /httpboot for ironic ports that aren't marked as PXE-capable is confusing. I get the point that it's best to let iPXE choose, but wouldn't it be better to let it chose from ports that are truly PXE capable?07:18
rpittaugood morning ironic! o/07:24
opendevreviewMerged openstack/ironic master: Update known issue for iDRAC Swift firmware update  https://review.opendev.org/c/openstack/ironic/+/84949508:23
opendevreviewAija Jauntēva proposed openstack/ironic bugfix/20.2: Update known issue for iDRAC Swift firmware update  https://review.opendev.org/c/openstack/ironic/+/84963308:27
opendevreviewAija Jauntēva proposed openstack/ironic stable/yoga: Update known issue for iDRAC Swift firmware update  https://review.opendev.org/c/openstack/ironic/+/84963408:28
opendevreviewDmitry Tantsur proposed openstack/bifrost master: CI: use openstacksdk from source if it's cloned  https://review.opendev.org/c/openstack/bifrost/+/84956308:36
opendevreviewMark Goddard proposed openstack/ironic stable/wallaby: Improve dependency resolution in driver-requirements.txt  https://review.opendev.org/c/openstack/ironic/+/84949108:41
timeuHi Ironic, I am trying to use baremetal port groups for setting up LACP on a baremetal node. If understand the documentation correctly if I don't set a MAC address on the port group, it should use the one from the neutron port (VIF). However for some reason the cloud-init network-data contains a random MAC address and not the VIF one which cause the node to not acquire a08:49
timeuDHCP lease. I guess the workaround is to use configdrive or explicitly set a MAC address but I was wondering if I do something wrong08:49
* dtantsur has no exposure to LACP. mgoddard ^^^08:56
dtantsurarne_wiebalck: polling for your interest. I may have a case for "active steps" downstream, specifically for firmware upgrades. Is it something that still interests you?08:56
mgoddardtimeu: I would expect the metadata to match the neutron port's MAC. Although I set the port group's MAC to one of the member ports (the one that is 'force up' if PXE booting over the bond)08:57
timeuyeah that was also my expectation reading the docs but it does not seem to do it. maybe it's not supported in the Train version that I try it on. I will try to set the MAC address to the first interface and see if it sets it. One other question tough the baremetal should be able to access the metadata service if it can reach the OS controllers ?09:01
timeufor some reason we also can't get the LACP bond working with NetworkManager. It always starts the bond and after a while it removes the slave and starts all over again. It works if I uninstall NetworkManager and install network-scripts in the image09:04
opendevreviewDmitry Tantsur proposed openstack/bifrost master: Wait for cleaning to be over when enrolling  https://review.opendev.org/c/openstack/bifrost/+/84965209:15
opendevreviewDmitry Tantsur proposed openstack/bifrost master: CI: use openstacksdk from source if it's cloned  https://review.opendev.org/c/openstack/bifrost/+/84956309:15
mgoddardtimeu: baremetal should be able to reach metadata with OVS, assuming there is a non-bonded network up09:52
mgoddardtimeu: if you just have a bond, use configdrive09:52
mgoddardtimeu: if you are using OVN, you can't get metadata currently09:53
timeumgoddard thanks for the info. Whhat's the technical reason why it doesn't work with bonded network ? 09:54
mgoddardtimeu: chicken and egg - you need the metadata to be able to bring up the bond09:54
timeuah yeah of course 09:54
timeumakes sense09:54
timeuI think in our case it worn't work anyways because we are using Cisco SDN + their distributed metadata agents09:55
mgoddardtimeu: I have seen the bond cycling issue before. IIRC I had the node on the wrong VLAN, so no connectivity09:55
timeuconfig drives works also setting the MAC address on the port group fixed the DHCP issue09:55
timeuthe last remaining issue is why the bond does't stay up with NetwokManager 09:55
timeuI tried it with setting NM_CONTROLLED=no on the intefaces. that fixes the issue as soon as I remove it NetworkManager starts to remove the slaves for some reason 09:56
mgoddardsee my last comment09:56
mgoddardI have it working with NM09:56
timeuhmm we don't have vlans configured only access 09:57
timeuwe also have other nodes that work fine with NM and LACP 09:57
timeunot sure why these nodes have issues. That's with RHEL 8.4 and MLX 5 dual port nics09:57
opendevreviewDmitry Tantsur proposed openstack/bifrost master: Wait for cleaning to be over when enrolling  https://review.opendev.org/c/openstack/bifrost/+/84965209:58
opendevreviewDmitry Tantsur proposed openstack/bifrost master: CI: use openstacksdk from source if it's cloned  https://review.opendev.org/c/openstack/bifrost/+/84956309:58
timeuinterestingly once the bond is up and running once without NetworkManager. Starting it again with NetworkManager seems to work and stable10:08
opendevreviewDmitry Tantsur proposed openstack/ironic-lib master: json_rpc.client: log the URL and exceptions  https://review.opendev.org/c/openstack/ironic-lib/+/84966110:38
opendevreviewwaleed mousa proposed openstack/ironic-python-agent master: Update and configure Nvidia nics fw by ironic-python-agent  https://review.opendev.org/c/openstack/ironic-python-agent/+/56654410:45
opendevreviewDmitry Tantsur proposed openstack/bifrost master: CI: use openstacksdk from source if it's cloned  https://review.opendev.org/c/openstack/bifrost/+/84956310:53
opendevreviewDmitry Tantsur proposed openstack/bifrost master: [WIP] Wait for cleaning to be over when enrolling  https://review.opendev.org/c/openstack/bifrost/+/84965210:53
opendevreviewMerged openstack/ironic-inspector master: Removed blacklisted to blocked  https://review.opendev.org/c/openstack/ironic-inspector/+/84591911:11
dtantsurfolks, could I get a 2nd +2 on https://review.opendev.org/c/openstack/bifrost/+/849074 and https://review.opendev.org/c/openstack/bifrost/+/848741 please? They're quite important.11:37
dtantsuralso any reviews on https://review.opendev.org/c/openstack/bifrost/+/849563/ please11:37
iurygregorydtantsur, will look after the meeting =)11:43
dtantsurthx!11:43
opendevreviewDmitry Tantsur proposed openstack/bifrost master: Wait for cleaning to be over when enrolling  https://review.opendev.org/c/openstack/bifrost/+/84965211:56
opendevreviewRiccardo Pittau proposed openstack/ironic-python-agent stable/xena: Gather details about bond interfaces if present  https://review.opendev.org/c/openstack/ironic-python-agent/+/84886812:30
opendevreviewRiccardo Pittau proposed openstack/ironic-python-agent bugfix/8.1: Gather details about bond interfaces if present  https://review.opendev.org/c/openstack/ironic-python-agent/+/84886912:35
opendevreviewRiccardo Pittau proposed openstack/ironic-python-agent stable/victoria: Gather details about bond interfaces if present  https://review.opendev.org/c/openstack/ironic-python-agent/+/84899112:43
opendevreviewRiccardo Pittau proposed openstack/ironic-python-agent stable/wallaby: Gather details about bond interfaces if present  https://review.opendev.org/c/openstack/ironic-python-agent/+/84899012:50
iurygregorydtantsur, done =)13:00
opendevreviewMerged openstack/bifrost master: Stop passing ironic_url to ansible modules  https://review.opendev.org/c/openstack/bifrost/+/84874113:07
TheJuliakaloyan: so I believe pxe_enabled is more "where we put PXE instructions in for Neutron", but I believe we've always created links for all interfaces which links to a central configuration file for the machine. iPXE does get to choose, but if you see it attempting multiple interfaces, then it may be that it is not taking the order of what it boots from. I guess we'd need a video recording of it network booting to really 13:07
TheJuliaunderstand since it seems different from the ipxe behavior I'm used to13:07
opendevreviewJulia Kreger proposed openstack/ironic master: DNM: Anaconda CI changes  https://review.opendev.org/c/openstack/ironic/+/84958713:10
kaloyanTheJulia: I'm using iPXE, maybe I didn't state that clearly13:13
iurygregoryjust me or gerrit is really SLOW today?13:17
TheJuliawell, to load ipxe, unless you've replaced the firmware in the cards, your ultimately initiating PXE first for a small network boot program13:17
TheJuliaand then it continues booting from there13:18
opendevreviewJulia Kreger proposed openstack/metalsmith stable/victoria: Add linters job  https://review.opendev.org/c/openstack/metalsmith/+/84967413:26
opendevreviewMerged openstack/ironic-python-agent stable/yoga: Gather details about bond interfaces if present  https://review.opendev.org/c/openstack/ironic-python-agent/+/84874213:33
opendevreviewMerged openstack/ironic master: Fix markup typo in Redfish driver docs  https://review.opendev.org/c/openstack/ironic/+/84961713:33
opendevreviewJulia Kreger proposed openstack/ironic stable/victoria: CI: disable some jobs which  https://review.opendev.org/c/openstack/ironic/+/84967713:35
opendevreviewVerification of a change to openstack/bifrost master failed: Deprecate os_ironic_node_info  https://review.opendev.org/c/openstack/bifrost/+/84876513:38
opendevreviewVerification of a change to openstack/bifrost master failed: Reduce the usage of no_log in keystone code  https://review.opendev.org/c/openstack/bifrost/+/84911413:38
opendevreviewJulia Kreger proposed openstack/ironic master: Deprecate syslinux  https://review.opendev.org/c/openstack/ironic/+/84222913:51
TheJuliaiurygregory: dtantsur do we have a list of bugfix branches which can be closed out?14:03
iurygregoryTheJulia, yes!14:04
TheJuliawe have way too many ones with busted CI14:04
iurygregoryfor ironic the ones we care are 20.2 / 19.0 / 18.1 14:05
iurygregorythe others I will ask to be removed14:05
TheJuliaiurygregory: can you trigger abandons on patches on those branches which have not merged?14:05
iurygregorywill do after the downstream meeting14:06
TheJuliak14:06
opendevreviewVerification of a change to openstack/ironic master failed: Allocation candidates prefer matching name  https://review.opendev.org/c/openstack/ironic/+/84585614:08
TheJuliaI've abandoned mine that are standing out to me14:09
TheJuliaCI must hate me14:12
TheJulialooks stuff is now failing all over the place where stuff worked as early as a few hours ago14:12
TheJuliaiurygregory: just so we keep in mind, https://review.opendev.org/c/openstack/ironic-python-agent/+/848969 is a major version on next release14:18
opendevreviewVerification of a change to openstack/ironic stable/wallaby failed: Improve dependency resolution in driver-requirements.txt  https://review.opendev.org/c/openstack/ironic/+/84949114:20
iurygregoryTheJulia, tks for the heads-up14:26
TheJuliai guess both get major versions on next release14:35
kaloyanTheJulia: the part with loading iPXE as a NBP works fine. The trouble arises from the fact that first script that iPXE loads is the boot.ipxe which loop through all NICs on the machine, trying to chainboot from them. This is the point where it failed14:42
TheJuliakaloyan: uefi or bios mode?14:44
-opendevstatus- NOTICE: Due to an incident in our hosting provider, the tarballs.opendev.org site (and possibly other sites served from static.opendev.org) is offline while we attempt recovery14:53
opendevreviewVerification of a change to openstack/ironic master failed: Make anaconda non-image deploys sane  https://review.opendev.org/c/openstack/ironic/+/83470915:03
dtantsuro/15:08
kaloyanTheJulia: UEFI15:25
kaloyanand BTW, https://tarballs.opendev.org/openstack/ironic-python-agent/dib/files/ipa-centos9-master.kernel returns a 403 as well as the directory, any idea why?15:26
rpittaulong weekend here, bye everyone, see you on monday! o/15:37
opendevreviewVerification of a change to openstack/ironic-python-agent master failed: Drop support for instance netboot  https://review.opendev.org/c/openstack/ironic-python-agent/+/84896916:19
iurygregorymaybe we should hold rechecks RETRY_LIMIT shows we couldn't reach some mirrors16:25
JayFYeah, note the status posted above16:49
JayFif tarballs.opendev is down16:49
JayFnothing is gonna pass I think16:49
JayFIPA jobs will want to upload there in post (so we shouldn't merge even if we could)16:50
JayFand Ironic jobs will get a prebuilt ramdisk from there16:50
JayFkaloyan: I guess you missed the notice?  Notice(opendevstatus): NOTICE: Due to an incident in our hosting provider, the tarballs.opendev.org site (and possibly other sites served from static.opendev.org) is offline while we attempt recovery16:51
JayFkaloyan: sorry, hopefully it gets back online soon16:51
TheJuliakaloyan: hmm, I'm guessing with uefi it is enumerating instead of just starting with the interface it has first17:09
* TheJulia tries to remember the code path it takes and suspects that is the case because it gets all of that from the simple networking stack in uefi17:09
TheJuliaugh, until tarballs is back, our CI is toasty17:10
iurygregoryyeah18:07
kaloyanTheJulia: here's my boot.ipxe https://pastebin.com/JJex3ewv I didn't use a custom template19:54
opendevreviewJulia Kreger proposed openstack/ironic-tempest-plugin master: Undo maintenace state on protected node tests  https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/84975220:10
TheJuliakaloyan: hmm, Lucas's note there makes me wonder...20:24
stevebaker[m]good morning!20:47
TheJuliagood morning20:47
* TheJulia raises an eyebrow at https://8c10dee59207be00b60c-b45f7fdceccc06baa60c38e38ca87fcf.ssl.cf1.rackcdn.com/840006/3/check/ironic-tox-unit-with-driver-libs/7d40c4d/testr_results.html20:50
JayF> AssertionError: Expected 'sleep' to be called once. Called 8 times.20:53
JayFthat is ... strange20:54
TheJuliayeah20:55
TheJuliaI haven't looked at the code, but I could guess maybe a rare race condition in one of the tests20:55
TheJuliabtw, I rechecked a bunch of stuff impacted by the great tarballs.o.o outage20:56
opendevreviewVerification of a change to openstack/ironic master failed: Make anaconda non-image deploys sane  https://review.opendev.org/c/openstack/ironic/+/83470921:05
opendevreviewVerification of a change to openstack/ironic stable/wallaby failed: Improve dependency resolution in driver-requirements.txt  https://review.opendev.org/c/openstack/ironic/+/84949121:06
opendevreviewVerification of a change to openstack/ironic master failed: Allocation candidates prefer matching name  https://review.opendev.org/c/openstack/ironic/+/84585621:07
kaloyanTheJulia: I'm kinda new to IRC so I didn't get what Lukas's note was :/21:24
TheJuliaoh, https://pastebin.com/JJex3ewv which is what you shared21:26
TheJuliakaloyan: ^^21:27
TheJuliaI'd have to look at the UEFI code, but I believe we're just looking at a different pattern of behavior21:27
TheJuliawith UEFI vs non-uefi21:27
opendevreviewMerged openstack/ironic bugfix/18.1: CI: Removing ironic job queue  https://review.opendev.org/c/openstack/ironic/+/84352023:22
jandersgood morning Ironic o/23:37

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!