=== narindergupta is now known as narinderguptamac
babbageclunkwallyworld: Here's the fix for bug 1847278: https://github.com/juju/juju/pull/1071401:33
mupBug #1847278: [vsphere] Juju does not respect --constraints root-disk-source to select appropriate datastore per node <juju:In Progress by 2-xtian> <https://launchpad.net/bugs/1847278>01:33
wallyworldbabbageclunk: looking01:35
wallyworldbabbageclunk: +1, did you manage to get vsphere to play nicely to test?01:41
babbageclunkyup yup01:42
hpidcockwallyworld: I've run the BlockDeviceMatching change against AWS, GCE and OpenStack with no issues. Do you think a run against vSphere is in order?02:21
wallyworldgiven it broke once, it would be good if possible, just to be 10000% sure02:21
babbageclunkwallyworld: tried extending a disk with the new govmomi, I still get a permission error waiting for the task02:35
babbageclunkI'll check that it still works with the wait and push that change anyway02:35
wallyworlddamn, ok. bug time02:35
babbageclunkactually, do you think I should upgrade the lib?02:36
babbageclunk(I mean, push the change updating the dep)02:36
babbageclunkwallyworld: ^02:38
wallyworlddoes that help with the polling of the dsk size?02:38
wallyworldif no, then i'd only do it for 2.702:38
babbageclunkyeah, makes sense02:39
babbageclunkno difference for the extend disk task02:39
wallyworldright, the errperm thing02:40
wallyworldbut we still need to fogure out how to poll02:41
babbageclunkwallyworld: my plan is to get them to use the cli tool govc to dump details of a VM on the datastore of interest to see how the structure differs from ours02:42
babbageclunkhopefully that would show what we should be looking at instead02:43
wallyworldfingers crossed02:43
wallyworldand hopefully an upstream bug will get some info as well02:43
wallyworldmaybe post the workaround we are using and see if anyone upstream replies to that02:44
babbageclunkok - will do both of those02:48
babbageclunkneed to go for a run though02:48
thumpermore robust controller config: https://github.com/juju/juju/pull/1071504:09
kelvinliuwallyworld: should we allow charm to create more than extra service account or just one?05:07
wallyworld> 1 i think is the requirement05:08
kelvinliuwallyworld: ok, https://github.com/juju/juju/pull/10716 could u take a look this PR? thanks!05:08
wallyworldsure, just pushing up one, will look in 505:09
kelvinliuwallyworld: np05:09
wallyworldkelvinliu: here's a small k8s peer relation fix https://github.com/juju/juju/pull/1071705:21
wallyworldkelvinliu: so the PR supports 1 SA in the kubernetesResources but we need to support > 1, ie a list05:29
wallyworldi believe that ken has operators that need > 105:29
kelvinliuah, just saw u said ">" 1.. my bad eyes. changing to a slice05:31
wallyworld\o/ ty05:33
kelvinliuwallyworld: changed to a slice of sa in KubernetesResources.05:46
wallyworldok, looking05:46
wallyworldkelvinliu: +1 ty. we'll need to check with ken tomorrow to ensure it meets his needs05:51
wallyworldhpidcock: you saw i +1 your pr?05:52
kelvinliuwallyworld: thx,05:52
hpidcockwallyworld: yep just want to get this vsphere run done + add a vsphere test in05:53
wallyworld\o/ ty05:53
hpidcockI don't want this to break again haha05:53
wallyworldand oracle05:53
hpidcockalready tested oracle05:56
wallyworldkelvinliu: shouldn't mark the bug as fix committed until the PR lands :-)06:24
kelvinliuwallyworld: ok..06:26
wallyworldleave for now, but for next one...06:26
kelvinliuchanged 1847125 back to in progress already06:27
=== pal__ is now known as parlos
=== nammn_de is now known as nammn_de_
nammn_de_any easy way to enforce upgrade steps run? Like a matcher whatever version I deploy to just run them again?08:26
manadartAnyone able to review https://github.com/juju/juju/pull/10697 ?10:09
achilleasamanadart: looking10:19
achilleasamanadart: or jam : If a relation unit used to have an ingress entry in its settings and when I try to refresh the network info I can't find one anymore, should I just delete the old entry from the settings?11:03
achilleasa(same question about egress-subnets)11:04
manadartachilleasa: I'm a bit light on this particular area, but I can't think why it would go away unless the machine NICs actually changed...11:06
manadartachilleasa: Actually, jump on a HO?11:07
achilleasamanadart: I was thinking something like machine rebooted and NIC is not there anymore11:07
nammn_de_achilleasa manadart any clue how to find out why "running machine configuration script" is taking ages, last log from `cloud-log-output.log` is "agent binaries downloaded succesfully. Its like half an hour for on lxd controller12:17
manadartnammn_de_: Blocked upgrade will do that.12:17
nammn_de_manadart: I did not run "upgrade-controller" how do I find out more to debug that?12:18
nammn_de_was just bootstraping12:18
manadartnammn_de_: Got a /var/log/juju/machine-0.log ?12:19
nammn_de_nope not created yet12:19
nammn_de_juju folder still empty, cloud-output seems to have finished. Trying to find oput where it is stuck12:20
bdx6wallyworld: <3 #1071714:29
mupBug #10717: gnome-vfs2: new changes from Debian require merging <gnome-vfs2 (Ubuntu):Fix Released by seb128> <https://launchpad.net/bugs/10717>14:29
bdx6heh - https://github.com/juju/juju/pull/10717/files14:29
bdx6^ big time thank you!14:30
=== bdx6 is now known as bdx
hmlachilleasa: actually… i do have an endpoint binding question for you.  why do we sometimes have the default EB specified, and sometimes not.  It’s not clear to me in these tests i’m fixing14:56
achilleasahml: test link? are they bundle-related tests?14:57
hmlachilleasa:  it’s all over, state, deploy (apiserver)14:58
hmlachilleasa:  mostly i see it, because the changes i’ve made have caused the default empty string endpoint to be returned in some places it wasn’t expected.15:00
achilleasahml: I wonder whether they were written assuming MAAS...15:02
hmlachilleasa: i’ve hit code from 2013 now… so who knows at this poit15:03
achilleasabecause otherwise they would end up un network.DefaultSpaceName, right?15:03
hmlachilleasa:  HO?15:03
achilleasasure, meet you in daily15:03
rick_hbdx:  cool was going to send you that link but didn't know if you'd get bdx6 lol15:33
bdxchristmas came early this year15:36
bdxyou guys rock15:36
rick_hbdx:  trying15:38
rick_hbdx:  ty for the help in chasing that down, strange one15:38
bdxyeah, a sneaky one15:44
nammn_de_hml: manadart: regarding our discussion from before, the old code does not seem to restart the agents:15:57
nammn_de_hml manadart: could it be that the models dont run the upgrade steps? I have upgraded the controller which did execute it (given log), then run "upgrade-model" they did not seem to run it, is there an option to do it?16:05
nammn_de_the upgradesteps I mean16:05
hmlnammn_de_:  to confirm, the models you ran upgrade-model with updated their version numbers?16:14
nammn_de_hml: yes they did. Juju status is returning the new version number16:15
nammn_de_has it something todo with steps being defined as 2.7.0 and we run 2.7-beta or is that independent and should work regardless?16:15
hmlnammn_de_: there are a few different types of upgrade steps.  state and non (one other that shouldn’t be necessary)16:15
nammn_de_hml: reading the go doc i cannot make sense of the difference beween `stateUpgradeOperations` and  `upgradeOperations` Is former only run on the controller?16:17
hmlnammn_de_: the state ones is for updating the database on the controller in one go.16:17
hmlnammn_de_: looking at the other init changes… you might need the other, non state version16:18
hmlnammn_de_: so stepsFor27() instead of stateStepsFor27()16:18
nammn_de_hm: I added to this slice as well and try it again, fi thats what you meant16:19
nammn_de_hml ^16:19
hmlnammn_de_: check out upgrades/steps_24.go and steps_245.go16:20
nammn_de_hml: I see, there actually seems to be a diff, should have looked closer thanks!16:21
nammn_de_hml: but we don't have a kind of graph/code doc how upgrades are handled, right? Like which comes first, where we set the log and so on (?)16:30
hmlnammn_de_: we don’t have a doc, the slices are executed in order.  depending on the order of the slice func.  can you expand on “where we set the log”, i’m not following16:31
nammn_de_hml: no that works perfect, I misswrote, sry. I mean lock :D16:32
hmlnammn_de_: those pieces happen automagically as part of  upgrades.  no additional locks should be necessary??16:33
nammn_de_hml: yes, I assumed that. Just wanted to know more for background understanding16:33
hmlnammn_de_: :-)16:34
nammn_de_hml: your tipp worked like a charm, thanks 🦸‍♀️!16:34
achilleasahml: if I have a machine instance do you know how can I update its addresses in the machine doc?16:36
achilleasa"SetDevicesAddresses" changes docs in other collections16:37
hmlachilleasa:  huh, let me look at something then16:37
hmlachilleasa:  SetMachineAddresses()?16:38
hmlachilleasa:  though it might be a set all…. not update16:39
achilleasadoh! how did I miss that? thanks!16:39
hmlachilleasa: there are any methods with address in machine.  hahaha  :-)16:39
manadartachilleasa: Note the difference between SetMachineAddresses (machine agent sourced) and SetProviderAddresses (from provider).16:54
achilleasayes, already stumbled on that :D16:54
achilleasaI need the latter16:55
manadartI'm not sure we really need MachineAddresses on that doc, but we can audit once things aren't in such flux.16:55
nammn_de_manadart: regarding the adding a upgrade step for space model config value, is to set for existing models the default value?17:03
manadartnammn_de_: Yes. If 2.7 assumes it's there, we need to set it for older models that are upgraded instead of bootstrapped anew.17:05
nammn_de_manadart: got it, thanks!17:05
pmatulis1hml, hi. when creating openstack image metadata for juju you need to supply a region. where does that value come from? is that from the region defined via 'juju add-cloud'? is it related to a "project" within openstack?17:46
hmlpmatulis1:  both.  they should be the same.17:46
hmlpmatulis1:  you can have multiple regions in the o7k cloud.  and in juju17:47
hmlpmatulis1:  openstack17:47
hmla shorthand of sorts i’ve seen17:47
pmatulis1ohh nice17:48
rick_hk8s, o7k, a11y, wheeee17:48
pmatulis1right right nice17:48
hmlor i18n17:48
hmlthat one really saves you17:48
pmatulis1what is it for? sorry17:48
hmlrick_h: ty!  long type.17:53
pmatulis1and a11y ?17:53
rick_hthat's my keyboard practive for the day17:53
rick_he.g. https://a11yproject.com/17:54
pmatulis1ah ha17:54
pmatulis1hml, so when i do 'openstack project list' i get back 'admin' and 'service'. r u (heh) saying i need to use one of those with 'juju metadata generate-image'?17:55
rick_hpmatulis1:  not projects, regions. Think us-east-1, us-east-2, etc17:56
hmlpmatulis1:  no, check your novarc file for an OS_REGION17:56
rick_hpmatulis1:  you can have lots of projects, those are kind of "per user" tenants17:56
rick_hpmatulis1:  but then you would deploy into your project resources from 1+ regions of the cloud (boston and chicago) or the like17:56
pmatulis1right, but hml i thought said project and region should be the same above. i must have misunderstood17:57
hmlpmatulis1: no, there are different.17:57
hmlpmatulis1: they17:58
hmlpmatulis1:  you need to match what’s in the cloud definition for your juju o7k cloud, when creating the image metadata17:59
hmlit’ll do an exact string match17:59
pmatulis1hml, ok, but there is no verification for that variable with add-cloud right?18:00
pmatulis1so i need to make sure it's correct18:01
hmlpmatulis1:  there is some versionification… to check that the URL is available18:01
hmlpmatulis1:  you can do add-cloud from the OS_ env var which should help18:01
pmatulis1hml, yeah, i'm working with microstack and stuff is not as normal18:02
pmatulis1it prolly should be though18:02
pmatulis1i don't have OS_REGION in my env18:02
hmlpmatulis1:  what does the juju cloud definition look like?18:02
pmatulis1hml, i decided to go cowboy and use 'localhost' since microstack is local. but in the end i've spoken to the developer and we're going to get OS_REGION hardcoded to 'microstack'18:06
pmatulis1and expose that to the user18:07
hmlpmatulis1:  what happens if you leave the region blank?18:07
hmlon both the simplestreams and the juju cloud def?18:07
pmatulis1hml, i haven't tried that18:08
pmatulis1hml, juju command says "ERROR image region must be specified"18:13
pmatulis1also, add-cloud doesn't work without that region line18:14
hmlpmatulis1:  rgr18:14
pmatulis1https://bugs.launchpad.net/microstack/+bug/1847649 <-- hml18:20
mupBug #1847649: Generating image metadata for Juju could be clearer in terms of Region <MicroStack:New> <https://launchpad.net/bugs/1847649>18:20
hmlpmatulis1:  makes me wonder if anything else is different for add-cloud etc.18:28
nammn_de_rick_h: wallyworld: as we were talking before. One of you might wanna take a look at this small one?18:48
rick_hnammn_de_:  sorry sure thing18:50
nammn_de_rick_h: No worries, gonna go now anyway. Btw my networ works again with all the beta upgrades :D18:51
rick_hnammn_de_:  yay!18:51
rick_hnammn_de_:  have a good night18:51
thumperhmm... another day where I forget to close IRC19:47
thumpermorning team19:47
hmlmorning thumper19:48
rick_hmorning thumper20:16
xavpaiceanyone know if there's a way to export the config of an application in a format compatible with "juju config mediawiki --file path/to/myconfig.yaml"?20:32
rick_hxavpaice:  if you do the --format=yaml is it not acceptable?20:41
xavpaicethat gives me the same output as 'juju config $application', but if I pipe that to a file, make an edit, then juju config $application --filename theyaml.yaml, it's a totally different layout20:43
xavpaiceI mean, export-bundle, edit, then deploy, should be fine in most cases, just wanting an app config I can 'borrow' from one place and put in another20:43
rick_hxavpaice:  yea, it should/would be that what you output you can accept20:44
xavpaicejuju config output is really good because it shows all the defaults, and the help, etc - but I can't use that to configure an app, needs a bunch of processing first20:45
rick_hxavpaice:  yea, understand.20:46
rick_hxavpaice:  the bundle path is closer but still not going to be right, as you say.20:46
rick_hxavpaice:  almost want a juju config --values-only20:46
rick_hor something20:46
xavpaiceyeah, exactly20:46
rick_hxavpaice:  so not something we currently have but I'm open to a bug on that. In general we've been looking at places where we output one thing but then don't accept it as input20:47
xavpaiceanyway, just a thought - if it was there already, would be handy to know about it, but I guess it's not so I'll just copy/paste out of a bundle20:47
rick_hxavpaice:  this falls under that umbrella for sure20:47
xavpaicecool - will scribble up a wishlist bug20:47
hpidcockhttps://github.com/juju/juju/pull/10721 2.6 to develop merge23:53

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!