[08:32:44] godog, herron, shdubsh o/ ES on logstash1009 alarmed, an OOM killed it, just re-ran puppet to bring it up again :( [08:36:28] elukey: ack, thanks! will take a look shortly too [10:40:54] just came accross this https://github.com/puppetlabs/puppet-editor-services (launguage and debug server for puppet) may be usefull to vscode users, have not tested my self yet [10:46:28] neat [10:59:27] <_joe_> jbond42: I think it might already be integrated in the puppet vscode extension [11:01:08] _joe_: would make senses thx [11:02:02] <_joe_> I'll check later :) [11:04:31] cheers :) [11:54:18] ooh, yall are using vscode? [12:03:54] sure as hell not emacs [12:04:35] * jbond42 uses vim and has booked marked https://github.com/puremourning/vimspector/issues/332 to investigate later [13:11:16] <_joe_> addshore: i'm mostly using vscode nowadays, yes [13:11:51] I have free licences for Wikimedia for the devsense PHP plugin [14:22:56] anyone here feel like admitting to ganeti knowledge? [14:23:14] i removed all disks from the d-i-test instance, because i wanted to add different ones [14:23:25] but removing the last disk turned out to cause Complication [14:23:40] what's the error? [14:23:53] Disk template conversion and other disk changes not supported at the same time [14:24:20] https://phabricator.wikimedia.org/P13661 [14:26:24] not sure, it's probably simply not supported. but we have plenty of head room in Ganeti currently, I'd suggest you simply create d-i-test02 etc. for the test layout you need [14:27:05] well, the existing instance is now hosed, right? it has no disks, and i can't add any [14:27:34] (meaning - i could recreate the existing one) [14:28:34] sigh, yeah. the docs say "there is no support for conversions to or from the diskless template" [14:28:49] which is a lie, as removing the last disk of an existing instance 'converts' it to a diskless template [14:29:25] so i guess i nuked d-i-test and remake it [14:29:28] *nuke [14:29:36] yeah, that should work [14:29:52] Gotta love trapdoor functions. [14:30:02] but if we need different setups in parallel for different install tests, feel free to just add a few more [14:30:47] i should whine at iustin next time i see him. (i used to work on ganeti-dev, a lifetime ago) [14:31:23] moritzm: ok cool, good to know [14:42:19] kormat: if you add/change the name we might need to add it to some list as d-i-test is treated as a special case in some places [14:42:24] because of no puppet [14:43:10] and if you nuke it do it with the decom cookbook (it *should mostly* work on that special case too ) [14:43:32] Might be a good idea to go with something like d-i-test[0-9]+ then, just in case. [14:43:46] yeah if we are creating more than one for sure [14:46:01] volans: i didn't change the name, but i also didn't run the decom cookbook. so i guess we'll find out what happens :) [14:46:26] kormat: you destroyed it at ganeti level? [14:46:30] yeah [14:47:24] * volans moves kormat from the nice people list to the not-nice one [14:47:26] :-P [14:47:32] volans: i think `abort` is the right answer here: https://phabricator.wikimedia.org/P13661#75286 [14:48:03] let me check [14:48:08] it might not work as you think [14:48:18] and we might need to clean that IP manually anyway in netbox [14:49:27] kormat: so... this use case is not supported, abort is the right choice but will not create the vm, we need a patch if we want to support this use case [14:49:49] i'm happy to create the vm manually in ganeti too [14:49:54] to avoid the complications [14:51:03] might be the easiest [14:51:21] how hard could it be, anyway [14:51:34] "Words to run from, fast." [14:51:34] the ideal easy fix is to just add a flag to skip allocation, but because we do auto-sync ganeti to Netbox the allocated IP might be deleted in the meanwhile [14:51:46] https://netbox.wikimedia.org/virtualization/virtual-machines/27/ already doesn't exists anymore [14:52:08] and the old IP was deleted from netbox [14:52:10] so kormat [14:52:15] if you're stull at the prompt [14:52:19] just say go and continue [14:52:25] i aborted [14:52:39] ok then give me a sec to check one thing [14:52:42] and we can re-run it [14:52:57] searching for https://netbox.wikimedia.org/search/?q=d-i-test [14:53:01] we have the new IP [14:53:04] if you delete that one [14:53:14] you can re-run the cookbook and follow what it does [14:53:22] shiuld replace d-i-test from old IP to new IP [14:53:39] delete what now? [14:53:41] I don't think the d-i-test IP was used anywhere hardcoded [14:53:47] delete https://netbox.wikimedia.org/ipam/ip-addresses/7505/ [14:54:04] the pre-allocated IP [14:54:07] deleted [14:54:11] that the cookbook had generated [14:54:43] ok, and i re-run the cookbook now? [14:54:47] yes [14:55:00] pynetbox.core.query.RequestError: The request failed with code 400 Bad Request: {'address': ['Duplicate IP address found in global table: 2620:0:861:101:10:64:0:41/64']} [14:55:44] damn netbox cache... it's really badly managed... [14:56:17] ahhh no [14:56:18] sorry my bad [14:56:30] you have to delete https://netbox.wikimedia.org/ipam/ip-addresses/7506/ too, you did flag it as skip v6 DNS? [14:56:33] why? [14:56:56] (that's why was not coming up with the query for d-i-test) [14:57:21] ironically, trying to reduce potential complications [15:01:55] all started with not running the decom cookbook fwiw :-P [15:02:26] https://netbox.wikimedia.org/search/?q=d-i-test now shows 2 v4 addresses [15:03:12] yes indeed, the first one was allocated when it failed to allocate the v6 one, we need to add more logic to the cookbook to rollback pre-allocated IPs on failure [15:40:22] kormat: care enough to create a task for the above maybe? [15:40:23] :) [15:51:42] volans: i do not :) [17:17:40] really nice - https://fosdem.org/2021/schedule/event/amd_gpus/ [17:17:50] we are not the only ones :D