| TheJulia | 0700 UTC... yeouch. but I have thoughts. :( | 00:03 |
|---|---|---|
| TheJulia | Then again, that is.. PM... | 00:03 |
| opendevreview | Julia Kreger proposed openstack/ironic-specs master: WIP Add a shard key https://review.opendev.org/c/openstack/ironic-specs/+/861803 | 00:04 |
| TheJulia | johnthetubaguy: JayF: ^^ tossing up an idea to get it out of my head. I'm thinking a migration could be "if microvresion is valid upon running, and an evacuate occurs, then we might be able to actually set a shard key value from nova-compute into ironic | 00:05 |
| TheJulia | *ironic's response* TBD | 00:05 |
| TheJulia | or not... there are some options there, but I think we could go either way, but we will ideally want to make assignment auto-magic | 00:05 |
| TheJulia | if at all possible. | 00:06 |
| TheJulia | I'd love for us to clean things up eventually, and it could just be a command to execute but locking would be *fun* albeit not impossible. | 00:08 |
| * TheJulia goes now | 00:08 | |
| shukun | TheJulia: Could you please review this patch: https://review.opendev.org/c/openstack/ironic/+/857035 | 04:00 |
| rpittau | good morning ironic! o/ | 06:51 |
| arozman | Hello Ironic! | 07:36 |
| dtantsur | TheJulia, JayF, our devstack configures tempest tests for both partition and whole-disk images. However, testing partition images on BIOS is not possible. In the normal case (as seen in the ironic jobs), we skip the partition test, but in case of misconfiguration, it may be run. | 10:27 |
| opendevreview | Jacob Anders proposed openstack/sushy master: Improve resiliency of eTag handling https://review.opendev.org/c/openstack/sushy/+/856123 | 10:28 |
| dtantsur | janders: I can do UTC morning, just let me know how it turns out | 10:32 |
| opendevreview | Merged openstack/ironic-lib stable/yoga: CI: Zuul no longer respects queue param https://review.opendev.org/c/openstack/ironic-lib/+/860172 | 10:35 |
| opendevreview | Merged openstack/ironic stable/victoria: CI: Reduce overall test load https://review.opendev.org/c/openstack/ironic/+/860851 | 10:37 |
| opendevreview | Merged openstack/ironic stable/ussuri: CI: Reduce overall test load https://review.opendev.org/c/openstack/ironic/+/860852 | 10:37 |
| janders | dtantsur TY. Let's chat more with TheJulia when she's online and I will run proposed times by Moshe, too | 10:39 |
| dtantsur | janders: would that mean cancelling the one at 22UTC? | 10:40 |
| dtantsur | (I guess no because the scale topic interests stevebaker[m]) | 10:41 |
| dtantsur | (but it may cancel it for me specifically :) | 10:41 |
| * janders is looking at the schedule | 10:41 | |
| * janders is firing up WorldClock | 10:42 | |
| janders | dtantsur I agree, the 22UTC session would remain | 10:44 |
| janders | if we wanted to do Active Steps in UTC morning, that would have to trickle down to Thursday, right? (it's kind of too late to do it still today) | 10:45 |
| janders | is that an option, WDYT? | 10:45 |
| janders | I missed that earlier cause Wednesday night UTC is my Thursday anyway | 10:45 |
| dtantsur | it's totally an option for me (we're past UTC morning already:) | 10:47 |
| janders | OK! let's work it out before today's session starts | 10:50 |
| iurygregory | JayF, I've updated the ptgbot =) | 12:42 |
| TheJulia | I'm a zombie, but good morning | 12:43 |
| dtantsur | good zombie morning TheJulia | 12:48 |
| dtantsur | what do zombies drink instead of coffee? | 12:48 |
| TheJulia | Oatmeal | 12:48 |
| TheJulia | acceptable texture which provides the fuel to do the zombie shuffle | 12:49 |
| * TheJulia sips coffee and starts to wake up | 12:50 | |
| janders | TheJulia JayF dtantsur before we start I wanted to touch again on re-scheduling the Active Steps discussion to a more Israel-friendly time | 12:57 |
| janders | re-pasting earlier idea to make things easier: | 12:58 |
| janders | We originally planned for 0700hrs UTC ( https://etherpad.opendev.org/p/ironic-antelope-ptg#L188 ) but the current schedule is 2200hrs UTC ( https://etherpad.opendev.org/p/ironic-antelope-ptg#L56 ). The new time doesn't suit him, that's 1am in his timezone. Do you think we still have room to move it? The original proposed time would work, | 12:58 |
| janders | alternatively 2000hrs UTC could maybe be considered, although it is getting late for the folks in Europe. What are your thoughts? | 12:58 |
| JayF | janders: and that was moved at the request of dtantsur /o\ | 13:00 |
| dtantsur | was it? Oo | 13:00 |
| JayF | janders: I'll bring it up, see what folks think. Worst case we just do a short meeting of the minds outside of the context of ptg | 13:00 |
| JayF | or iurygregory | 13:01 |
| JayF | I think it was iurygregory then, sorry dtantsur | 13:01 |
| dtantsur | 22UTC could not be something I requested.. yeah, Iury maybe | 13:01 |
| iurygregory | we talked that 22 would be a good time for APAC to participate, since in the past we had the same sessions in EU/America TZ and America/APAC TZ | 13:09 |
| janders | iurygregory makes sense | 13:15 |
| janders | in this case I wonder if we should add a repeat session on Active steps, say 0800hrs UTC (pedning Moshe/Mellanox availability)? | 13:16 |
| janders | this way we don't need to stuff around with the existing schedule at last minute | 13:16 |
| janders | let me know if this seems like a good idea and I can reach out to Moshe | 13:16 |
| dtantsur | if it's not earlier than 8UTC, it's doable | 13:33 |
| JayF | janders: if you all want to meet at 0800 UTC; you can. Please take good notes as that is basically maximally-worst timezone time for me. | 13:35 |
| janders | dtantsur JayF thanks! Checking with Moshe/Mellanox, if it works let's try. We can treat it as a repeat session. | 13:37 |
| * TheJulia reads depressing news and sighs | 13:55 | |
| dtantsur | sorry to hear TheJulia! | 13:58 |
| TheJulia | eh, just another attempt to ban my existence... at a national level. | 13:58 |
| iurygregory | TheJulia, ouch =( sorry to hear *hugs* | 13:59 |
| dtantsur | oh of course... these people will never forward their energy to something useful and peaceful, will they? | 13:59 |
| iurygregory | probably not | 14:00 |
| TheJulia | I really wish they would! | 14:00 |
| iurygregory | arne_wiebalck, if possible in the ironic session we are talking about raid | 14:07 |
| iurygregory | can you join? | 14:07 |
| dtantsur | janders: any response re the active steps slot? | 14:24 |
| johnthetubaguy | TheJulia: FWIW, I was wondering about evacuate as possible move nova-compute operation, it almost fits. I guess we need to decide if moving between nova-compute-ironic services is a one off transition thing, or a regular rebalance my shards thing. If its a one off, I like not changing the API I guess. | 14:40 |
| JayF | johnthetubaguy: we were talking in PTG: it'll either need to be rebalancable and/or nova compute needs to be able to handle "N" shard keys per n-c (so you can preprovision ala cassandra vnodes) | 14:42 |
| johnthetubaguy | I think one of the assumptions we had on the Nova side was "this doesn't change, for a given node uuid"... but I think if the service is down, you should be able to move it again. | 14:44 |
| johnthetubaguy | sorry, been in magnum land today, and nova/kolla land yesterday, so not had chance to loop back, I might try write something down before I forget it all | 14:45 |
| JayF | We agreed to accept that limitation in the same way that other drivers accept it; it's hard for me to imagine someone being upset about hooking up `migrate` or `evacuate` if we kept the model properly | 14:56 |
| janders | dtantsur nothing from Moshe so far. Based on response or lack of, we can decide if we have a follow-up session during the 2200hrs "main" active steps session. It would be small-ish anyway. Would this work? | 15:00 |
| janders | I will set the alarm clock - will be in the 2200hrs session :) | 15:00 |
| dtantsur | yeah | 15:02 |
| dtantsur | I'll try not to fall asleep :) | 15:02 |
| JayF | I'm sorry about that :( It's very difficult to get everyone booked at good times | 15:02 |
| janders | JayF I think we're doing well - thanks for organising everything | 15:03 |
| JayF | dtantsur: I need to step away, I might be back in time to start session on inspector merge; please start without me if I'm not back | 15:07 |
| dtantsur | when did we agree to start? | 15:07 |
| dtantsur | I was out on a call | 15:07 |
| TheJulia | 10 after | 15:07 |
| stevebaker[m] | good morning | 15:09 |
| opendevreview | Kirill proposed openstack/ironic master: add vnc console https://review.opendev.org/c/openstack/ironic/+/860689 | 15:33 |
| * TheJulia blinks | 15:40 | |
| TheJulia | zoom crashed.. again | 16:06 |
| TheJulia | zoom why do you hate me | 16:06 |
| iurygregory | zoom hate us all | 16:07 |
| rpittau | bye! o/ | 16:08 |
| opendevreview | Kirill proposed openstack/python-ironicclient master: add support of new graphical console interface https://review.opendev.org/c/openstack/python-ironicclient/+/860688 | 16:16 |
| hjensas | TheJulia: Raised bug/rfe on port list https://storyboard.openstack.org/#!/story/2010373 - I found the old bug on performance/scale issue and linked to that, this new bug also brings up my wishes related to networking-baremetal agent scaling. | 16:34 |
| TheJulia | hjensas: ack, thanks | 17:01 |
| hjensas | TheJulia: Looks like we could do the conductor_group filtering in a similar way we do owner and project filtering? Would probably be a good thing for me too look into. | 17:03 |
| TheJulia | hjensas: yes, that definitely would be good to do in the networking-baremetal service | 17:04 |
| TheJulia | hjensas: likely just a configuration key for networking-baremetal to decide how it will query Ironic. | 17:04 |
| hjensas | TheJulia: yes, I'll start working on adding add_port_filter_by_node_conductor_group in db api. | 17:07 |
| TheJulia | oh yeah | 17:07 |
| TheJulia | that... would be needed too | 17:08 |
| TheJulia | if you beat me to making the query/result set a join cool | 17:08 |
| TheJulia | I'm thinking this may be a case where working with our existing object model for getting the result set may just be a "bad idea" fwiw. | 17:08 |
| opendevreview | Kirill proposed openstack/python-ironicclient master: add support of new graphical console interface https://review.opendev.org/c/openstack/python-ironicclient/+/860688 | 17:50 |
| opendevreview | Julia Kreger proposed openstack/ironic-specs master: WIP Add a shard key https://review.opendev.org/c/openstack/ironic-specs/+/861803 | 18:08 |
| TheJulia | JayF: johnthetubaguy: ^ So I don't think nova-computes can handle n number of shard keys. The conundrum becomes what happens if a new shard key is created... and well... a new nova-compute would be needed realistically. And I think that is okay. this is a case where automatic scalability was a problem to support overall, and explicit decisions on scaling need to be made on the nova side. From inventory management, we can | 18:21 |
| TheJulia | just... handle it somehow. And I think likely some sort of tool to stamp existing baremetal nodes with a key and then a nova operator being able to do a rolling migration makes sense to me | 18:21 |
| janders | dtantsur TheJulia JayF no response from Moshe, so I don't think we'll be doing alternate Active Steps session. Let's stick with the original plan, we can have further discussions outside PTG. | 21:19 |
| JayF | Alright. Thank you for coming even though it's a rough time | 21:19 |
| JayF | This is the 15 minute warning for the final Ironic PTG session at UTC 2200-2300 | 21:46 |
| opendevreview | Julia Kreger proposed openstack/ironic-specs master: WIP Add a shard key https://review.opendev.org/c/openstack/ironic-specs/+/861803 | 21:55 |
| TheJulia | okay, I think the idea is out of my head | 21:56 |
| TheJulia | oh, we would need a tool to do the shard key setting, it could be the base migration though | 21:56 |
| TheJulia | well, no | 21:56 |
| TheJulia | online migration, it can be done while running | 21:56 |
| JayF | I admire your capacity to do such things after 4-5 hours of video meetings | 21:57 |
| dtantsur | I guess I need to have a fresher brain to understand the difference with conductor groups | 21:58 |
| TheJulia | dtantsur: I will write that then! | 21:58 |
| TheJulia | or at least, expand upon it | 21:58 |
| dtantsur | ++ would be great | 21:58 |
| dtantsur | I also hope it won't be completely parallel to conductor groups, but I'll wait for the write-up (and enough capacity to read it) | 21:59 |
| TheJulia | oh absolutely not parrallel | 22:00 |
| * TheJulia feels like "#ironic now bikeshedding verbs" | 22:12 | |
| dtantsur | go for it! | 22:12 |
| dtantsur | stevebaker[m]: side note: I'd love to know how much in common we have between metal3/openshift and what you're doing | 22:35 |
| dtantsur | our timezone overlap is disastrous :) | 22:35 |
| dtantsur | maybe janders could sync with you.. or maybe there is a slot which is not deep night for either of us | 22:36 |
| JayF | dtantsur: just putting this here -- Ironic remote pxe support. e.g. Ironic using a remote agent on a machine to configure http/tftp/pxe environment *there* | 22:40 |
| * JayF wonders if one could abuse webdav for that | 22:40 | |
| dtantsur | on the machine itself? or just splitting away a "PXE as a Service"? | 22:40 |
| JayF | I'm saying if you had an nginx in a node-facing position, it could run an agent/webdav/be-sshed-to to configure pxe/tftp/http boot | 22:41 |
| JayF | PXEaaS makes it sound much more difficult but probably is more truthful to the difficulty | 22:41 |
| dtantsur | you're inventing a shared directory :) | 22:41 |
| dtantsur | we'll need locks, so NFS is actually better IMO | 22:42 |
| opendevreview | Merged openstack/ironic master: Add support auth protocols for iRMC https://review.opendev.org/c/openstack/ironic/+/857035 | 23:10 |
| stevebaker[m] | dtantsur: yeah now that I actually understand the space I'm ready for a proper engagement with metal3. I could do a late night for your morning | 23:32 |
| opendevreview | Julia Kreger proposed openstack/ironic-specs master: WIP Add a shard key https://review.opendev.org/c/openstack/ironic-specs/+/861803 | 23:34 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!