Monday, 2017-06-05

*** thorst has joined #openstack-powervm00:02
*** k0da has quit IRC00:07
*** thorst has quit IRC00:18
*** thorst has joined #openstack-powervm01:14
*** thorst has quit IRC01:19
*** svenkat has quit IRC01:36
*** edmondsw has joined #openstack-powervm01:41
*** edmondsw has quit IRC01:45
*** thorst has joined #openstack-powervm01:59
*** thorst has quit IRC02:26
*** thorst has joined #openstack-powervm03:26
*** thorst has quit IRC03:31
*** thorst has joined #openstack-powervm03:33
*** thorst has quit IRC03:35
*** thorst has joined #openstack-powervm04:05
*** thorst has quit IRC04:22
*** edmondsw has joined #openstack-powervm05:17
*** edmondsw has quit IRC05:21
*** chhavi has joined #openstack-powervm06:08
*** thorst has joined #openstack-powervm06:18
*** thorst has quit IRC06:23
*** chhavi has quit IRC06:38
*** edmondsw has joined #openstack-powervm07:05
*** edmondsw has quit IRC07:09
*** thorst has joined #openstack-powervm07:20
*** thorst has quit IRC07:24
*** k0da has joined #openstack-powervm07:44
*** thorst has joined #openstack-powervm08:20
*** chhavi has joined #openstack-powervm08:22
*** chhavi_ has joined #openstack-powervm08:32
*** chhavi has quit IRC08:33
*** thorst has quit IRC08:40
*** edmondsw has joined #openstack-powervm08:53
*** edmondsw has quit IRC08:57
*** thorst has joined #openstack-powervm09:37
*** thorst has quit IRC09:41
*** edmondsw has joined #openstack-powervm10:41
*** edmondsw has quit IRC10:45
*** svenkat has joined #openstack-powervm11:23
*** chhavi_ has quit IRC11:34
*** thorst has joined #openstack-powervm11:51
*** efried has quit IRC12:09
*** efried has joined #openstack-powervm12:21
*** mdrabe has joined #openstack-powervm12:27
*** edmondsw has joined #openstack-powervm12:55
*** chhavi_ has joined #openstack-powervm12:57
*** jwcroppe has quit IRC13:05
*** esberglu has joined #openstack-powervm13:08
*** jpasqualetto has joined #openstack-powervm13:29
*** jwcroppe has joined #openstack-powervm13:32
*** tjakobs has joined #openstack-powervm14:40
*** kriskend has joined #openstack-powervm14:42
*** kriskend has quit IRC14:57
esbergluthorst: efried: edmondsw: PowerVM support matrix
esbergluWould like a sanity check there. Some of the feature descriptions were iffy15:02
edmondswtx esberglu15:04
edmondswefried should attach/detach vol be "partial" at this point?15:05
efriednope, missing.15:05
edmondswefried missing? even with the SSP stuff merging?15:05
efriedThat's ephemeral.  This is referring to volume.15:05
edmondswoh, true15:06
efriedI haven't gotten all the way through, but ephemeral may be implicitly supported.15:06
edmondswefried line 981... I can't tell if that includes ephemeral or not15:12
efriedgetting there...15:12
edmondswI love the notes for 718 :)15:14
edmondswand the fact that it has wtf in the name15:14
efriededmondsw 981, yeah, I don't think that's what we have.  I think that's, like, swift/ceph/whatever.15:15
edmondswswift is object storage, not block15:16
edmondswefried I was confused as to why it even mentioned swift there15:16
edmondswefried I assumed it was only to elaborate on the "alternative" to this line15:17
efriedI think our ephemeral is covered by 107815:17
efriedesberglu Posted responses.  Definitely want thorst to take a look, and it would be nice to get adreznec's eye on it as well.15:18
esbergluefried: Okay I'll wait until have a chance to take a look to update15:19
efriedesberglu Let's not cross off stuff in the etherpad until it's merged.15:23
efriedthorst do you know offhand if is cpu_time_ns considered necessary to claim complete support for get-guest-info?
thorstI don't think so?15:26
efriedCause we don't even have that OOT.15:26
thorstbut not for certain15:26
efriedthorst The notes list it ("cummulative CPU execution time") without any provisos or caveats, so just based on that, it would seem so.  We don't even have that OOT.15:27
efriedNot even sure if that's available to us.15:27
efried...without accumulating from PCM from the birth of the instance, which I don't see us doing.15:27
thorstwe can but we can't15:28
thorstif it ever live migrates, the PCM data is removed15:28
thorstwhich should be the same with libvirt...15:28
edmondswdo we already have resize support in?15:29
thorstotherwise, yeah, we may actually be able to work15:29
thorstedmondsw: no - I already indicated it in the review15:29
edmondswthorst tx, I hadn't seen your comments yet15:29
thorstliterally just posted15:29
efriedI think someone's going to have to poke through the other drivers to see if/how they support that cumulative CPU thing.15:32
efriedthorst Does PCM even give us a total-since-inception for a VM?15:34
thorstI thought so?15:34
thorstunless it wipes it out on a VM restart15:34
thorstbut again, that'd be the same for libvirt15:34
efriedIf it doesn't, we've got no way to do this anyway, cause we wouldn't have a way to store it across compute process and/or REST server restarts.15:34
thorstlooking at libvirt code, it seems to me they have the same problem.15:35
efriedGuess we should at least raise a LP bug to look into this.15:35
*** dwayne_ has joined #openstack-powervm15:41
openstackLaunchpad bug 1695924 in nova-powervm "Cumulative CPU time not implemented" [Low,Confirmed]15:44
*** k0da has quit IRC15:44
*** chhavi_ has quit IRC15:58
*** kriskend has joined #openstack-powervm16:04
*** kriskend has quit IRC16:20
*** kriskend has joined #openstack-powervm16:20
*** chhavi has joined #openstack-powervm16:35
*** k0da has joined #openstack-powervm18:01
*** chhavi has quit IRC18:01
openstackgerritEric Fried proposed openstack/nova-powervm master: Performance improvements for Lifecycle events
efriedthorst adreznec edmondsw esberglu ^^ is ready for final reviews.  TIA.18:03
*** tjakobs has quit IRC18:25
*** k0da has quit IRC18:26
*** tjakobs has joined #openstack-powervm18:49
efriedthorst edmondsw Soooo... get_info is used in only two places in nova, and both of them only look at the state, nothing else.18:54
efriedIn another wrinkle, the docstring on the superclass method def says we're supposed to be looking up the instance by its name, not its ID, which we've never done.18:55
edmondswefried link?18:55
*** kriskend has quit IRC18:57
efriedvmware tries to get by UUID and then name:
edmondswefried we even have that comment in our docstring:
efriedWell, yes, we copied & pasted it, but didn't actually do that.19:00
edmondswyeah, the next line of the docstring says it's an obj, not a string19:00
efriedlibvirt does it by name19:00
efriedYeah, I'm not disputing that we get an Instance object - but the docstring for some reason says we're supposed to do the lookup by name, and the other drivers (so far) seem to be actually doing that.19:01
efriedI wonder if the docstring is moldy, and there's no need for it to be by name anymore, but when it got converted, the consumers only updated to index into the obj instead of changing over to lookup by UUID19:02
edmondswefried I think it must be19:02
efriedgit blame time.19:02
edmondswif we weren't getting an object, this would blow up:
edmondswI'm guessing the docstring is from before it was an object19:03
edmondswonce it's an object, why would they care what from the object we use to do the lookup?19:03
efriedyuh.  Looks like mikal missed it in I54b32a5020b0dbc21ba7156ed38ed188c483086b19:09
*** esberglu has quit IRC19:31
*** k0da has joined #openstack-powervm19:33
*** esberglu_ has joined #openstack-powervm19:33
efriedthorst edmondsw esberglu_  Following up: I'm considering not implementing those other fields.  If we leave it the way it is, the first time nova does anything that attempts to use them, our CI will blow up with AttributeError.  This will give us the opportunity to see how they're being used, and evaluate how we should implement them.19:43
edmondswefried did you ask any of the nova cores if they know of plans to use them?19:44
efriedLike, if the usages are scattered and few, we may wish to use quick props, like the OOT driver does today.19:44
efriedIf they're used densely all together, we may wish to get the whole LPAR wrapper.19:44
efriedIf the static-ish ones are used frequently from various places, we may wish to cache them for longer than an InstanceInfo lives.19:45
efriededmondsw I did not.  Kinda hinted at it to mriedem, but he didn't bite.19:45
efriedFrom what I know about how they're doing resource providers and allocations and such, I have reason to believe they'll never care about what we report.  They'll register those amounts in the db on spawn/resize and just consult those numbers to calculate how much the host has consumed.  Which is probably what they're doing today.19:46
edmondswefried might be good to ask, so that if there are some plans we can implement based on those and not worry about the CI breaking.19:47
edmondswefried But if we can't uncover any plans, I think I'm fine with what you suggested19:47
edmondswefried comments up on
*** esberglu_ has quit IRC20:07
*** esberglu has joined #openstack-powervm20:09
*** esberglu has quit IRC20:09
*** esberglu has joined #openstack-powervm20:10
*** edmondsw has quit IRC20:26
*** thorst has quit IRC20:28
*** thorst has joined #openstack-powervm20:28
*** thorst has quit IRC20:33
*** thorst has joined #openstack-powervm20:50
*** thorst has quit IRC20:54
*** dwayne_ has quit IRC20:58
*** svenkat has quit IRC21:19
*** thorst has joined #openstack-powervm21:23
*** edmondsw has joined #openstack-powervm21:28
*** edmondsw has quit IRC21:32
*** kriskend has joined #openstack-powervm21:42
*** jpasqualetto has quit IRC21:52
*** kriskend has quit IRC22:02
*** thorst has quit IRC22:08
*** thorst has joined #openstack-powervm22:09
*** thorst has quit IRC22:13
*** thorst has joined #openstack-powervm22:42
*** thorst has quit IRC22:42
*** tjakobs has quit IRC22:43
*** jwcroppe has quit IRC22:53
*** thorst has joined #openstack-powervm22:55
*** thorst has quit IRC22:55
*** jwcroppe has joined #openstack-powervm23:24
*** thorst has joined #openstack-powervm23:26
*** jwcroppe has quit IRC23:28
*** k0da has quit IRC23:41
*** thorst has quit IRC23:42

Generated by 2.15.3 by Marius Gedminas - find it at!