/srv/irclogs.ubuntu.com/2017/12/06/#juju-dev.txt

* thumper loads up his controller01:30
thumperhopefully this won't actually kill my laptop01:30
thumperI'm trying to replicate the death spiral we see with some controllers with lots of models01:30
hmlthumper: ha, you’re brave?01:36
thumperwell... I'm up to 300 models and 60 total units01:37
thumperstarting to see rises in txns.log reads01:37
thumpersince I'm on an SSD, no doubt my profiling will be different01:37
thumperbut slowly adding stuff to see it get stressed01:37
hmlwallyworld: i pushed the pr changes, but there is something wrong with the tests, but real world does what it should on packages.01:37
hmlwallyworld: i’m EOD, look at tests in the morning01:37
wallyworldok, ty01:38
thumpercurrently: goroutine profile: total 16650001:57
thumperaxw: the new hook execution locking behaviour is 2.3 not 2.2.6 right?01:58
axwthumper: correct01:59
* thumper nods02:00
* thumper slowly adds more models and units02:00
thumperperhaps unsurprisingly, my machine is starting to get a little sluggish02:05
thumperwallyworld: I've got a problem, got a minute?02:21
wallyworldsure02:21
thumper1:1 HO02:21
mwhudsonan overloaded controller _and_ hangouts? brave man :)02:25
axwwallyworld: the state.Unit.ConigSettings method adds defaults, the application method does not02:42
axwConfigSettings*02:42
wallyworldaxw: yeah, just discovering that02:42
wallyworldjust otp with tum02:43
cmarsi'm running into this bug with the manual provider: https://bugs.launchpad.net/juju/+bug/173658203:08
mupBug #1736582: Cannot bootstrap manual provider with Juju 2.2.6 <juju:New> <https://launchpad.net/bugs/1736582>03:08
cmarsis there a workaround?03:08
thumperhmm...03:33
thumperwhen I restarted the controller03:34
thumperit would have hit the provider asking for machines for each model in very short order03:34
thumperstarting all the provisioners in parallel has the potential to DDOS the provider03:35
thumperwhich is what we did for lxd03:35
wallyworldoops, we should fix that03:49
wallyworldthumper: i would expect model migrations to copy across charms - it that something that is done out of band? or is it in the export yaml?03:50
wallyworldi'm guessing out of band?03:50
wallyworldor is that just resources that gets streamed across03:51
babbageclunkwallyworld: charm binaries are done out of band - they're not in the export yaml04:08
wallyworldbabbageclunk: thanks. so that means we potentially have a situation for a short time where an application has an orphaned charn reference. do we not mark the model as ready until all blobs are copied?04:12
babbageclunkwallyworld: that's right - the binaries (agents, charms and resources) are transferred in the import phase, and the model isn't marked as active until the success phase (when all the agents are switched over)04:17
wallyworldbabbageclunk: thanks! i need to update a test then due to to a change i'm making04:17
babbageclunkwallyworld: no worries (he says, too late)04:38
wallyworldnever too late04:38
* babbageclunk still has time, it's never to late to change his mind, it's - never - too late - to change his mind04:38
wallyworld_axw: for when you are free https://github.com/juju/juju/pull/817905:15
axwsure, just finishing up my PR then will take a look05:15
wallyworld_no rush05:15
axwwallyworld: https://github.com/juju/juju/pull/818005:55
wallyworldaxw: looking. i'm just pushing a change to fix a feature test failure05:55
axwokey dokey05:55
wallyworlddone05:56
wallyworldaxw: our test had a comment saying "this is wrong" referring to the source attr value, but it wasn't fixed, just a comment left behind05:56
axwwallyworld: lol :/05:57
wallyworldyeah05:57
axwwallyworld: so we don't know who owns the juju namespace?05:58
wallyworldaxw: no, "they" are trying to find out05:58
wallyworldno one knows05:58
axwok05:58
wallyworldaxw: did you use the latest version of my charm? i added extra config attrs which are set in the pod spec05:58
axwwallyworld: nope, and I hacked up the charm a bit to set the spec05:59
axwmaybe yours does that now?05:59
wallyworldyeah, since yesterday05:59
wallyworldi'll test once all this lands05:59
=== frankban|afk is now known as frankban
jamaxw: did you see bug #1729880, someone is wanting to run a patched version of Juju but it is in a multi-architecture deploymente09:53
mupBug #1729880: juju 2.2.4 and 2.2.6 actions dissappear when state is changed from running to complete <juju:Fix Committed by ecjones> <https://launchpad.net/bugs/1729880>09:53
jamI thought you had a script that would allow cross-compiling and uploading alternative agent binaries.09:53
jamfwiw, all he really needs is the controller to be patched, he could use official binaries for the rest of the agents (I think).09:54
jambut probably our bug about an upgraded controller creating new units/machines with the *controller* agent version rather than the *model* agent version.09:54
axwjam: I didn't. stepping out shortly, but I'll be back later to respond09:54
axwjam: FYI, it's https://github.com/axw/juju-tools09:55
jamaxw: have a good night09:55
anastasiamacwallyworld: ping12:04
wallyworldyo12:09
anastasiamacwallyworld: pm'ed12:14
=== freyes__ is now known as freyes
=== frankban is now known as frankban|afk
=== frankban|afk is now known as frankban
balloonshml, can I get a +1? https://github.com/juju/juju/pull/818217:45
=== frankban is now known as frankban|afk
* hml looking18:05
hmlballoons: lgtm18:07
babbageclunkthumper: ping? Do you have a moment for some advice?21:17
thumperbabbageclunk: sure21:17
babbageclunkthumper: in 1:1?21:18
thumperyep21:18
wallyworldhml: pop into stand up HO? just want to touch base about the config stuff22:02
hmlwallyworld: on my way22:02

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