Tuesday, 2024-01-23

Sandzwerg[m]<TheJulia> "*or* haven't changed/updated/..." <- I to this day frequently look at the ironic state diagram, although less frequent than I used to. That does not mean that I have any preference on the format these diagrams in00:30
TheJuliaSandzwerg[m]: Truthfully, I really dislike the format of the diagram which gets generated.  But short of rewriting the code to assemble it... *shrugs*00:32
Sandzwerg[m]Yeah rewriting stuff can be a lot of effort. But I think apart from the state diagram I don't remember many diagrams in the ironic docs. I never looked at all the vendor drivers to be fair and mostly try to find what I know or assume to be there, so I might miss stuff compared to reading it from the start to the end00:43
opendevreviewMerged openstack/ironic master: Stop using a specific mirror in infra  https://review.opendev.org/c/openstack/ironic/+/90625601:24
KangieHi All, building out a new node provisioning / management system for our smallest HPC cluster has finally percolated to the top of my priority list. Ironic seems to fit most of what I need. Is there anything that I should know / read before I start deploying bifrost to do some initial testing, useful tips, etc?01:53
JayFWhat's the top line on the basic shape of your environment and requirements? What's your desired end state? and do you want any bells and whistles (e.g. fancy networking) 02:05
KangieHow much of the openstack stack are you going to sell me?02:16
Kangie:p02:16
KangieOK, what I'm looking at _for now_ is ~10 nodes + a head node02:17
Kangiebells and whistles I'd like would be the ability to flexibly deploy different types of node images (which I think is already possible out-of-the-box/via plugins02:18
Kangieand being able to define networking config, automagically (in particular bridged networks) would be particularly desirable here02:18
Kangieand useful if/when we decide to scale out to a larger cluster02:18
KangieI guess the current baseline is Nvidia's Bright cluster manager. We'll probably end up building new authentication infra so we don't need that feature02:19
Kangiethe rest is basically sugar around defining node configurations / networks / etc and the automagic to make that work within bright's 'box'02:19
KangieI have a personal interest in trying to use OCI images to define our compute node config(s), but we're already used to managing chroot images so that could also be adopted02:20
KangieNote that I _am_ open to being sold on more of the stack, so if there'a already a nice authentication plugin that we can use as a source of truth for users and groups that's fine, but we can also do openldap/389ds/etc independently which might end up being desirable.02:22
Kangieotherwise nothing too fancy, though longer term I'm hoping to have less 'hard' walls between our clusters so we can use software to magically redefine nodes and drop them on the right vlans for a different cluster (say via fancy switch config magic)02:23
Kangieoh and just to make things interesting we're now looking at adding a small pool of virtual login nodes to our next major cluster iteration02:25
Kangie(not part of this cluster, but something I'll need to someday manage...)02:37
JayFif anything, we're angled more towards the service provider direction: there are services that provide auth *for the cloud services* but not services that provide auth *for servers running in the cloud*03:18
JayFthat'd be a workload we'd be happy to run03:18
JayFif that makes sense03:18
JayFFancy networking in the sense I meant it is more: having Ironic and friends (neutron, is the only current option) coordinate switching networks to enhance security03:19
JayFso I can put a server on a dedicated provisioning network when provisioning, move it into "tenant A" network when provisioned by tenant A (or even have them pick), then move into a cleaning network to be cleaned safely03:19
JayFif you just need on-disk networking config once provisioned, again, that's in your workload -- Ironic will pass through a configdrive containing some metadata including a network-data.json which is read by several standard tools, including cloud-init and glean03:20
JayFand those tools usually do the initial setup like embedding ssh keys and converting that network metadata into whatever format $distro would want it in03:21
JayFbifrost crosses these lines a little more, I think, by setting up some of the configdrive pieces for you03:22
JayFand being a coordinator between ironic and some of your images03:22
JayFthe lines I'm drawing here are Ironic lines :) bifrost is ansible that is sometimes meant to blur those lines to make it easier03:22
JayFbut I figured, based on our previous chats in #gentoo-chat, you might appreciate knowing where the actual borders are underneath03:24
opendevreviewTakashi Kajinami proposed openstack/ironic stable/2023.2: Stop using a specific mirror in infra  https://review.opendev.org/c/openstack/ironic/+/90628903:40
opendevreviewTakashi Kajinami proposed openstack/ironic stable/2023.1: Stop using a specific mirror in infra  https://review.opendev.org/c/openstack/ironic/+/90629003:42
opendevreviewTakashi Kajinami proposed openstack/ironic master: Clean up removed services from devstack options  https://review.opendev.org/c/openstack/ironic/+/90632603:48
opendevreviewTakashi Kajinami proposed openstack/networking-generic-switch master: Clean up removed services from devstack options  https://review.opendev.org/c/openstack/networking-generic-switch/+/90632903:53
KangieYeah, That's sort of where I'm at now.03:58
Kangieneutron might be interesting though, how good is it at reprovisioning nodes for, say, tenant B?03:59
KangieIf each of our clusters can take advantage of that functionality it might meet some of my flexibility goals.03:59
Kangiethough I guess for us in the physical world that might need to be done via switch port reconfiguration, or we just have one big management network down the line.04:01
Kangie(sorry, got side tracked with lunch)04:01
TheJuliaThe model is generally you have a central neutron, and you pull the machines back into a provisioning network by re-attaching them to perform the OS deploy, then you put them back on the tenant's requested network automatically04:39
KangieI'll read some docs and try and wrap my head around that04:55
KangieProbably not needed for the initial 10-node pilot anyway04:55
TheJuliaoh, definitely not04:55
TheJuliaAnyway, my cats are signaling it is time for sleep.04:56
TheJuliaGoodnight!04:56
KangieGoodnight!04:56
opendevreviewTakashi Kajinami proposed openstack/ironic master: Use local mirror to download images for anabonda  https://review.opendev.org/c/openstack/ironic/+/90634306:15
songwenping_hi, team, any body can tell me where the "$TEMPEST_BAREMETAL_MIN_MICROVERSION" define pls?06:36
tkajinamsongwenping__, https://github.com/openstack/ironic/blob/master/devstack/lib/ironic#L58507:29
tkajinamit's empty unless you (or a job) override it07:29
rpittaugood morning ironic! o/07:53
tkajinamrpittau, o/08:10
rpittauhey tkajinam :)08:10
tkajinamhttps://review.opendev.org/c/openstack/ironic/+/906289 was merged and uefi-resfish-vmedia job is green so CI in sushy/sushy-tools is also green now \o/08:12
tkajinamwould be nice if we can merge these backports, too http://mirror-int.ord.rax.opendev.org:8080/rdo/centos9-master/component/network/3e/4a/3e4a3848963f64a0fd752a5b440290c8a04d764d_fb5748ee08:16
-opendevstatus- NOTICE: all new logins to https://review.opendev.org are currently failing. investigation is ongoing, please be patient08:54
jrossersongwenping_: you can also use codesearch to find things like that https://codesearch.opendev.org/?q=TEMPEST_BAREMETAL_MIN_MICROVERSION09:00
iurygregorygood morning Ironic09:33
songwenpingtkajinam, how can i start a job to override it?10:39
tkajinamsongwenping, add you override under vars.devstack_localrc https://github.com/openstack/ironic/blob/master/zuul.d/ironic-jobs.yaml#L3510:41
tkajinamyou can find other environments being overridden there10:41
songwenpingtkajinam, how the ironic tempest plugin skip the test for microversion, i cannot find the "$TEMPEST_BAREMETAL_MIN_MICROVERSION" set in old releases.10:51
opendevreviewMerged openstack/ironic stable/2023.2: Stop using a specific mirror in infra  https://review.opendev.org/c/openstack/ironic/+/90628911:03
tkajinamsongwenping, as long as the common helper from tempest is used, no tests with microversion requirements would be skipped without min/max_micro_version11:14
tkajinamI don't know which tests and which branch you are referring to. We may need to these as well as the real reason of tests tests being skipped to dig int that11:15
iurygregorywe also set some tempest regex on some jobs, so maybe it doesn't run in the job you were looking11:21
iurygregoryjust a reminder Don't log out from review.opendev.org =) 12:20
opendevreviewTakashi Kajinami proposed openstack/sushy-tools master: Improving nova search filter  https://review.opendev.org/c/openstack/sushy-tools/+/89613012:56
opendevreviewDmitry Tantsur proposed openstack/ironic-lib master: Split common qemu-img functions from disk_utils  https://review.opendev.org/c/openstack/ironic-lib/+/90637112:58
opendevreviewDmitry Tantsur proposed openstack/ironic-lib master: Split common qemu-img functions from disk_utils  https://review.opendev.org/c/openstack/ironic-lib/+/90637113:00
opendevreviewTakashi Kajinami proposed openstack/ironic master: Use local mirror to download images for anaconda  https://review.opendev.org/c/openstack/ironic/+/90634313:11
opendevreviewDmitry Tantsur proposed openstack/ironic-python-agent master: Drop usage of run_as_root  https://review.opendev.org/c/openstack/ironic-python-agent/+/90637513:24
opendevreviewMerged openstack/sushy-tools master: Add Python 3.10 and 3.11 to supported versions  https://review.opendev.org/c/openstack/sushy-tools/+/90621613:28
opendevreviewDmitry Tantsur proposed openstack/ironic master: WIP auto-discovery  https://review.opendev.org/c/openstack/ironic/+/90637613:56
opendevreviewMerged openstack/ironic stable/2023.1: Stop using a specific mirror in infra  https://review.opendev.org/c/openstack/ironic/+/90629014:01
opendevreviewMerged openstack/ironic master: Clean up removed services from devstack options  https://review.opendev.org/c/openstack/ironic/+/90632614:01
TheJuliagood morning14:11
iurygregorygood morning TheJulia =)14:15
dtantsurLooking for more reviews for the new inspection stuff, especially the docs in https://review.opendev.org/c/openstack/ironic/+/904766 but also https://review.opendev.org/c/openstack/bifrost/+/896925 https://review.opendev.org/c/openstack/ironic/+/906308 and https://review.opendev.org/c/openstack/ironic/+/90631414:44
iurygregoryI've approved https://review.opendev.org/c/openstack/ironic/+/904766 14:55
dtantsurthx!15:01
opendevreviewMerged openstack/ironic master: PXE configuration guide for unmanaged inspection  https://review.opendev.org/c/openstack/ironic/+/90476615:30
* JayF will be OOO today; out sick15:43
dtantsurget better JayF!15:46
opendevreviewDmitry Tantsur proposed openstack/ironic master: Add node auto-discovery support for in-band inspection  https://review.opendev.org/c/openstack/ironic/+/90637615:50
opendevreviewDmitry Tantsur proposed openstack/ironic master: Add node auto-discovery support for in-band inspection  https://review.opendev.org/c/openstack/ironic/+/90637616:04
*** mmalchuk_ is now known as mmalchuk16:06
dkingdtantsur: That's a good point about the authentication. I suppose that some people will not want to expose that nor expect kubernetes to have that access. In that case, it seems like it could make sense to create a separate endpoint which internally checks the health of the software without exposing secure information.16:14
JayFSuch an endpoint would also be a DoS attack vector.16:19
JayFYou'd have to have it limited to healthcheckers via IP blocking or policy anyway16:20
JayF(at least assuming it was wired to give semi-real-time information)16:20
opendevreviewMerged openstack/ironic master: Flip require_managed_boot to True for the new agent inspection  https://review.opendev.org/c/openstack/ironic/+/90631416:26
dtantsurJayF: could be using the cached hash ring16:27
opendevreviewDmitry Tantsur proposed openstack/ironic master: Add node auto-discovery support for in-band inspection  https://review.opendev.org/c/openstack/ironic/+/90637616:35
dtantsuryour mention of DoS made me add a warning here ^^16:35
opendevreviewDmitry Tantsur proposed openstack/ironic master: Add node auto-discovery support for in-band inspection  https://review.opendev.org/c/openstack/ironic/+/90637616:51
TheJuliaThis sounds a lot like the healthcheck vs healthcheck v2 stuffs17:05
rpittaugood night! o/17:15
TheJuliadtantsur: fyi, tried to get it all out of my head at this point: https://review.opendev.org/c/openstack/ironic-specs/+/906324/1/specs/approved/fix-vmedia-boot-config.rst  It might just be easiest to try and have a meeting of the minds at some point to discuss further17:21
opendevreviewMerged openstack/ironic master: Account for nodes with the same BMC hostname in inspection lookup  https://review.opendev.org/c/openstack/ironic/+/90630817:59
dtantsurTheJulia: will read asap (but not today) - thanks!18:21
*** osmanlicilegi is now known as Guest1618:22
opendevreviewJulia Kreger proposed openstack/ironic master: typo: fix a typo in api configuration  https://review.opendev.org/c/openstack/ironic/+/90639418:26
iurygregoryIf anyone has time for review today https://review.opendev.org/c/openstack/ironic/+/903379 =)18:28
opendevreviewMerged openstack/ironic master: add default conductor group capability  https://review.opendev.org/c/openstack/ironic/+/85570519:17
opendevreviewJulia Kreger proposed openstack/ironic master: Add HTTP versions of network boot interfaces  https://review.opendev.org/c/openstack/ironic/+/90096519:29
-opendevstatus- NOTICE: The Gerrit service on review.opendev.org will be offline momentarily for a restart, in order to attempt to restore OpenID login functionality19:36
iurygregoryyay19:36
TheJuliayay19:37
-opendevstatus- NOTICE: OpenID logins for the Gerrit WebUI on review.opendev.org should be working normally again since the recent service restart20:03
opendevreviewMerged openstack/ironic master: Drop rootwrap support  https://review.opendev.org/c/openstack/ironic/+/90483620:14
opendevreviewMerged openstack/sushy-tools master: Remove translation sections from setup.cfg  https://review.opendev.org/c/openstack/sushy-tools/+/90621721:50
opendevreviewMerged openstack/ironic master: typo: fix a typo in api configuration  https://review.opendev.org/c/openstack/ironic/+/90639421:54
opendevreviewMerged openstack/ironic master: Disable legacy RBAC policy by default.  https://review.opendev.org/c/openstack/ironic/+/90200921:58
opendevreviewJulia Kreger proposed openstack/ironic-tempest-plugin master: DNM/WIP: Detect misconfig and navigate  https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/90643223:14
opendevreviewJulia Kreger proposed openstack/ironic-tempest-plugin master: WIP/DNM: Scope is enforced  https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/90643423:16
opendevreviewKyrylo Romanenko proposed openstack/python-ironicclient master: Add test for driver raid property list  https://review.opendev.org/c/openstack/python-ironicclient/+/48048923:38
opendevreviewKyrylo Romanenko proposed openstack/python-ironicclient master: Add test for driver raid property list  https://review.opendev.org/c/openstack/python-ironicclient/+/48048923:45

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