=== cpaelzer__ is now known as cpaelzer | ||
otubo | blackboxsw: sorry for not responding in time, timezone differences :-( there's also a BZ related to this: https://bugzilla.redhat.com/show_bug.cgi?id=1772505 | 09:39 |
---|---|---|
ubot5 | bugzilla.redhat.com bug 1772505 in cloud-init "swapon fails with "swapfile has holes" when created on a xfs filesystem by cloud-init" [High,Assigned] | 09:39 |
otubo | blackboxsw: just in case you want to attach that to the commit message as well | 09:39 |
rharper | Odd_Bloke: thanks | 09:44 |
Odd_Bloke | rharper: I'm not sure for what, but you're welcome. ;) | 14:49 |
blackboxsw | ahosmanMSFT: note that I'm not seeing byteswapping behavior on gen2 Standard_B1s from dmi data. The reason being that I think walinuxagent package is writing out a proper /sys/devices/virtual/dmi/id/product_uuid file with the right byte ordering. | 17:18 |
blackboxsw | ahosmanMSFT: I *think* a walinuxagent fix may have been introduced as a workaround for this byte swap issue I think? | 17:18 |
blackboxsw | or maybe I need a different vm --size that does exhibit the problem. | 17:19 |
blackboxsw | when I check dmi data file it has correct byte ordering | 17:19 |
ahosmanMSFT | blackboxsw, maybe, I haven't checked waagent/cloud-init in a bit. FYI I'll be switching from provisioning soon. | 17:19 |
ahosmanMSFT | let me get you a repro | 17:19 |
blackboxsw | Odd_Bloke: https://github.com/cloud-init/ubuntu-sru/pull/89/files is my SRU test for azure byte-swapping behavior. Though currently I haven't found a reproducing instance type | 17:22 |
blackboxsw | thanks ahosmanMSFT | 17:22 |
Odd_Bloke | blackboxsw: Shouldn't that be in the [Test Case] section of the SRU template? | 17:23 |
ahosmanMSFT | blackboxsw: So this will take a little bit to set up. We'll need to get two 1.65 TIP nodes, then upgrade one to 1.8. The two nodes would be part of a single availability set that we can pin a VM to. We'd exclude the 1.8 node to force the VM to be deployed to the 1.65. Then un-exclude the 1.8 node, kill the 1.65 node to force service healing to move the VM to the 1.8 node | 17:23 |
Odd_Bloke | (We had also talked about forcing an incorrect UUID for testing, which might be the best route if you can't find a reproducing instance type.) | 17:23 |
Odd_Bloke | ahosmanMSFT: What's "TIP" here? | 17:24 |
ahosmanMSFT | Testing In Prod | 17:24 |
Odd_Bloke | Aha, thanks. :) | 17:24 |
ahosmanMSFT | Here's the full convo I had with support | 17:25 |
ahosmanMSFT | https://www.irccloud.com/pastebin/j4ASsCWL/convo | 17:26 |
ahosmanMSFT | The UUID is changed in Azure backend | 17:26 |
Odd_Bloke | Right, OK. | 17:26 |
blackboxsw | roger. yeah ok that may be a bit more involved than I can reproduce easily via az commands. so I'll just trigger fake uuid swap and ensure it works | 17:27 |
blackboxsw | easy enough for me to manipulate what is cached in cloud-init datasource to a swapped value | 17:27 |
blackboxsw | ok have an appt, back in a bit | 17:27 |
blackboxsw | thanks | 17:27 |
ahosmanMSFT | No worries, support tested it and I verified there results | 17:29 |
=== johnsonshi_ is now known as johnsonshi | ||
meena | https://github.com/canonical/cloud-init/pull/161 when will be the right time to merge this? | 21:34 |
Odd_Bloke | meena: Oh, I've been meaning to comment there. Because you weren't the original author, we need them to sign off on the CLA. | 21:44 |
Odd_Bloke | *If you're thinking that that's extremely annoying, then we are in agreement.) | 21:45 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!