wallyworld | hpidcock: https://github.com/juju/juju/pull/11360 | 01:30 |
---|---|---|
wallyworld | babbageclunk: lgtm, ty | 01:39 |
babbageclunk | wallyworld: thanks! | 01:39 |
hpidcock | wallyworld: oops sorry, was lying down with a headache | 02:45 |
wallyworld | all good, hopr you're ok | 02:45 |
tlm | does anyone know why we start certificates 7 days before now and not 5 minutes in the past ? | 03:45 |
* wallyworld shrugs | 03:45 | |
tlm | ¯\_(ツ)_/¯ | 03:46 |
kelvinliu | wallyworld: these two PRs for adding storage support for non stateful app, https://github.com/juju/charm/pull/308 https://github.com/juju/juju/pull/11362 could u take a look? | 06:36 |
parlos | Good Morning Juju! | 07:44 |
stickupkid | manadart, got this for CR https://github.com/juju/juju/pull/11355 | 09:05 |
achilleasa | stickupkid: can you do a quick sanity check on a 2.7 -> dev merge PR https://github.com/juju/juju/pull/11363? | 09:05 |
manadart | stickupkid: Yep; will look. | 09:05 |
stickupkid | achilleasa, done | 09:06 |
=== parlos is now known as parlos_afk | ||
manadart | stickupkid: https://github.com/juju/juju/pull/11364 | 09:14 |
=== parlos_afk is now known as parlos | ||
stickupkid | manadart, https://media.giphy.com/media/OCgTKYSVnf7iM/giphy.gif | 09:34 |
stickupkid | manadart, i have no idea what i was trying to say "Subnet defines the nes" - nes what? nintendo nes? what was I doing? | 09:35 |
stickupkid | manadart, quick ho? | 09:44 |
manadart | stickupkid: OMW. | 09:44 |
=== parlos is now known as parlos_afk | ||
achilleasa | jam: I have replied to your comments in 11341 and pushed two additional commits (the second deals with that odd c-tor for the relation resolvers); can you take another look? | 10:23 |
=== parlos_afk is now known as parlos | ||
stickupkid | manadart, this looks wrong to me, I would expect this just check for if != 1, as you want to know why it wasn't removed https://github.com/juju/juju/blob/develop/api/spaces/spaces.go#L196-L198 | 10:37 |
flxfoo | hi all | 10:37 |
flxfoo | It is me again | 10:37 |
flxfoo | :) | 10:37 |
manadart | stickupkid: Yes. | 10:38 |
stickupkid | flxfoo, hi o/ | 10:38 |
flxfoo | So while trying to instanciate `juju add-machine` (with constraints works fine)... so two things... first juju returns 11 pending pending bionic instance "31c020be-c4b9-440d-aaf0-d2c1b33ea218" has status BUILD, wait 10 seconds before retry, attempt 8, and just loop , until I do a `juju remove-machine X --force" | 10:39 |
flxfoo | so then after a little delay, list of instances is empty ... on rackspace side though... I have a list of 11 instances with the same name (difference IDs) which will listed as building (90% at most) after some time though the instances starts to appear as Ready | 10:40 |
flxfoo | So I suspect something wrong with juju, maybe it does not receive the right "return" so It keeps call for server instantiation... | 10:41 |
flxfoo | stickupkid: hi :) | 10:42 |
stickupkid | flxfoo, so this comes from https://github.com/juju/juju/blob/develop/provider/openstack/provider.go#L1240 | 10:43 |
stickupkid | flxfoo, we're trying to provision a machine, but we're not getting back one that meets the constraints OR the provider doesn't have enough resources for the machine to be created | 10:44 |
stickupkid | flxfoo, I've seen this locally when testing with multipass+microstack and it was the latter case for me | 10:45 |
flxfoo | stickupkid: this is the only place right? | 10:49 |
stickupkid | flxfoo, let me check | 10:49 |
flxfoo | stickupkid: on rackspace side (as I said) each instance have the same name, but different IDs (which make sense), but juju return always the same ID to retry... why do I have several instances created on rackspace though? | 10:51 |
stickupkid | flxfoo, that's correct https://paste.ubuntu.com/p/pk76srRk37/ | 10:51 |
stickupkid | flxfoo, that I don't know | 10:52 |
stickupkid | flxfoo, if you think you've got a reproducer then I would create a bug - https://bugs.launchpad.net/juju/+bugs | 10:53 |
=== parlos is now known as parlos_afk | ||
flxfoo | stickupkid: no idea yet, just trying to put things together | 10:53 |
flxfoo | stickupkid: do you know if I explicitely pass `-n 1` that would make a difference? | 10:55 |
=== parlos_afk is now known as parlos | ||
stickupkid | flxfoo, tbh, I wouldn't know personally | 10:56 |
flxfoo | stickupkid: Would you know how a single `add-machine` could endup looping in endless server creation? | 11:04 |
stickupkid | flxfoo, not getting back one that meets the constraints OR the provider doesn't have enough resources for the machine to be created | 11:04 |
=== parlos is now known as parlos_afk | ||
=== parlos_afk is now known as parlos | ||
stickupkid | manadart, haha, there are so many issues in remove space, trying to work it out now | 11:50 |
stickupkid | in the cmd/remove.go | 11:50 |
=== parlos is now known as parlos_afk | ||
stickupkid | anyone know how to test that something was written to a cmd/ctx log? | 12:13 |
stickupkid | manadart, that was a pig | 12:40 |
stickupkid | https://github.com/juju/juju/pull/11365 | 12:40 |
=== parlos_afk is now known as parlos | ||
=== parlos is now known as parlos_afk | ||
manadart | achilleasa: 11356 looks good here, but is still marked as a draft. | 14:23 |
achilleasa | manadart: marked it as a draft bec it needs a rebase/force-push once the relation-created one lands | 14:51 |
manadart | achilleasa: OK. I've approved it. | 14:51 |
achilleasa | manadart: tyvm | 14:52 |
=== parlos_afk is now known as parlos | ||
=== parlos is now known as parlos_afk | ||
=== parlos_afk is now known as parlos | ||
achilleasa | is there a juju-idiomatic way for accessing controller config options inside state? I was thinking of having the facade fetch them and pass them as arguments to the state method I am working on but I am wondering whether we use a different pattern | 15:58 |
=== parlos is now known as parlos_afk | ||
=== parlos_afk is now known as parlos | ||
=== parlos is now known as parlos_afk | ||
=== parlos_afk is now known as parlos | ||
achilleasa | rick_h_: do you think 640k for charm data and 512k for uniter data are reasonable defaults for the quota limits? 640k ought to be enough for everyone, right? ;-) | 17:39 |
achilleasa | note that the operator will still be able to set the limits to 0 and bypass the quota checks if they are feeling yolo | 17:40 |
rick_h_ | achilleasa: hmmm, ideally folks will never know/hit this. I'd prefer to start with bigger defaults like 512k for juju and 1 or 2M for units? | 17:41 |
achilleasa | rick_h_: OK, though 2M seems too generous to me. The absolute max (combined) is 16M | 17:44 |
rick_h_ | achilleasa: right, personally I'd almost default to the max. I mean we store charms with 1GB of resources/etc | 17:45 |
rick_h_ | achilleasa: it seems silly to make folks open the faucet more ever and only close it if they hit issues/care | 17:45 |
achilleasa | rick_h_: but those end up in blob store, right? | 17:46 |
rick_h_ | achilleasa: but I can go wit some limit to start I guess. We know most controllers in the wild don't have > 3 models and most models are 3-5 applications | 17:46 |
rick_h_ | achilleasa: yes, this is true | 17:46 |
rick_h_ | achilleasa: but it's on the same disk as the rest of mongodb | 17:46 |
achilleasa | rick_h_: we can also do a 14M/1.5M split as the default and allow operators to fine-tune it if they need to | 17:47 |
achilleasa | my concern is basically that if we set it too high, charm authors will end up abusing it | 17:48 |
achilleasa | to store binary blobs or something | 17:48 |
achilleasa | (logs) | 17:48 |
rick_h_ | achilleasa: yea, that's why I'm ok with going roomy, but not all the way. I feel like the 512k 2M means we're talking 2.5mb per unit on the machine MAX. Realistically not all your charms will be abuses. | 17:48 |
rick_h_ | achilleasa: but we definitely need a metric about size of this collection so we can add it to graphana and if there's a disk usage rise track it directly to this change | 17:49 |
rick_h_ | grafana doh | 17:49 |
=== parlos is now known as parlos_afk | ||
=== parlos_afk is now known as parlos | ||
* babbageclunk waves | 19:37 | |
ventura | are u using IRC for project development and discord for app usage? or migrating everything to discourse? | 19:38 |
rick_h_ | ventura: irc for real time chat but discourse for async/news/published details | 19:45 |
rick_h_ | most simple charm is ubuntu heh, just get a machine and setup the charm environment | 19:49 |
ventura | rick_h_: turn off/on feature flags on backend for mobile clients | 19:55 |
ventura | if it possible to create a git repos with config changes, juju would allow to easily git-revert bad flags with the benefit of always keep track of changes | 19:58 |
flxfoo | hi again, :) | 20:00 |
rick_h_ | ventura: that does some work but as Juju is more than configuration management there's a lot more to "changes" than flipping config flags. | 20:00 |
rick_h_ | ventura: there's resources, or binaries provided to run/use. There's actions, that trigger administrative functions like adding users, backing up db, etc. There's relations, that instruct application to pass details about themselves back and forth. | 20:00 |
rick_h_ | ventura: a lot more moving parts and "live system" than can be easily git commit/rollback | 20:01 |
flxfoo | so I can confirm that when a machine is created (juju add-unit / add-machine) the process on the provider finishes (takes time but finish)... Except that there is more than one machine created with the same name (different IDs)... | 20:02 |
flxfoo | I had to do a `juju remove-machine XX --force`, because after 11 instances... it is too much :) | 20:03 |
flxfoo | I think there is something as well here, where rackspace needs time to allocate resources, and the frequency where a new server will be spawned... | 20:03 |
ventura | rick_h_: i mean "a simple charm that shows something to my manager allow using juju" :-) | 20:04 |
ventura | TL; DR: we lost all machines configs during Black Friday due Bolsonaro Bug (i.e. crazy day-light saving time changes in Brazil) | 20:04 |
flxfoo | I don't know If I could put a delay like 30 minutes intead of 10s | 20:04 |
flxfoo | Anyone on the fact that calling one `juju add-unit/add-machine` would end up create multiple instance with the same name (different IDs) ? | 21:27 |
thumper | flxfoo: I'm not sure I understand what you mean | 21:29 |
thumper | wallyworld: https://github.com/juju/lumberjack/pull/1 | 21:29 |
flxfoo | @thumper: when I perform a `juju add-unit` juju loop saying that instance is in state BUILD, retrying... probably due to rackspace lagging... but then a few minute later I have another one instamce created (with the same name different id) and juju report the same message (with different ID) | 21:33 |
flxfoo | until I `juju remove-machine --force` that just goes on | 21:33 |
thumper | flxfoo: that definitely sounds like a bug | 21:34 |
flxfoo | after removing on juju side... all the instances go from buidling state to ready state | 21:34 |
thumper | perhaps due to the slow nature in rackspace | 21:34 |
thumper | we don't see it on our other openstacks | 21:34 |
flxfoo | yeah that sounds very much something linked to rackspace.. | 21:35 |
flxfoo | feww weeks ago that was not doing that for sure | 21:35 |
wallyworld | thumper: lgtm | 23:20 |
babbageclunk | thumper: if the password we use to connect to mongo for juju-restore is always oldpassword from controller machine 0, what happens if machine 0 has gone away? ie, the controller has machines 1, 2, 3? | 23:22 |
thumper | what does the juju-db plugin do? | 23:23 |
thumper | we should use that | 23:23 |
babbageclunk | should we be connecting as a different user, so we can use the oldpassword from the machine we're on? | 23:23 |
babbageclunk | ooh good call - looking | 23:23 |
babbageclunk | ah, ok - it uses statepassword and the tag | 23:25 |
babbageclunk | I think that might have been the problem with how we were doing it before - trying with tag but oldpassword, not statepassword | 23:25 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!