Friday, 2016-09-02

*** seroyer has quit IRC00:09
*** seroyer has joined #openstack-powervm00:22
*** thorst has joined #openstack-powervm00:24
*** kriskend_ has joined #openstack-powervm01:46
*** kriskend has quit IRC01:46
*** esberglu has joined #openstack-powervm01:58
*** esberglu has quit IRC02:08
openstackgerritDrew Thorstensen (thorst) proposed openstack/nova-powervm: WIP: Send full fabric to VIOS slot data for LPM
*** thorst has quit IRC02:45
*** thorst has joined #openstack-powervm02:46
*** thorst has quit IRC02:54
*** wangqwsh has joined #openstack-powervm03:32
*** tjakobs has joined #openstack-powervm03:42
*** thorst has joined #openstack-powervm03:52
*** tjakobs has quit IRC03:57
*** thorst has quit IRC03:59
*** kriskend_ has quit IRC04:04
*** wangqwsh has quit IRC04:13
*** tjakobs has joined #openstack-powervm04:44
*** tjakobs has quit IRC04:48
*** thorst has joined #openstack-powervm04:56
*** thorst has quit IRC05:04
-openstackstatus- NOTICE: The logs volume is full causing jobs to fail with POST_FAILURE. This is being worked on, please do not recheck until notified.05:10
*** ChanServ changes topic to "The logs volume is full causing jobs to fail with POST_FAILURE. This is being worked on, please do not recheck until notified."05:10
*** ChanServ changes topic to "This channel is for PowerVM-related development and discussion. For general OpenStack support, please use #openstack."05:36
-openstackstatus- NOTICE: Space has been freed up on the log server. If you have POST_FAILURE results it is now safe to issue a 'recheck'05:36
*** thorst has joined #openstack-powervm06:02
*** kotra03 has joined #openstack-powervm06:02
*** kotra03 has quit IRC06:04
*** thorst has quit IRC06:09
*** kairo has joined #openstack-powervm06:11
*** kotra03 has joined #openstack-powervm06:23
*** thorst has joined #openstack-powervm06:36
*** thorst has quit IRC06:44
*** k0da has joined #openstack-powervm07:30
*** thorst has joined #openstack-powervm07:42
*** thorst has quit IRC07:48
*** thorst has joined #openstack-powervm08:46
*** thorst has quit IRC08:54
*** thorst has joined #openstack-powervm09:52
*** thorst has quit IRC09:58
*** kairo has quit IRC10:25
*** kairo has joined #openstack-powervm10:35
*** thorst has joined #openstack-powervm10:56
*** kairo has quit IRC10:58
*** kairo has joined #openstack-powervm11:02
*** thorst has quit IRC11:04
*** kairo has quit IRC11:06
*** kairo has joined #openstack-powervm11:21
*** thorst has joined #openstack-powervm11:57
*** tblakes has joined #openstack-powervm12:01
*** tblakes_ has joined #openstack-powervm12:02
*** tblakes has quit IRC12:06
*** tblakes_ is now known as tblakes12:07
*** kotra03 has quit IRC12:25
*** kairo has quit IRC12:25
*** kairo has joined #openstack-powervm12:25
*** kairo has joined #openstack-powervm12:25
*** apearson has joined #openstack-powervm12:43
*** dwayne_ has quit IRC12:49
efriedthorst, yt?13:08
thorstefried: yeah...just working on hardware requests.13:08
efriedReady to do some steppin'?13:08
thorstSR-IOV ones actually...the last ones13:08
*** tjakobs has joined #openstack-powervm13:08
efriednetwork_infos and crt_network_infos13:08
thorstcan I finish the SR-IOV hardware request?  I'm in the zone...13:08
thorstbefore I get mad about a different thing  :-)13:09
efriedSho.  But I found some very exciting and interesting stuff out.13:09
efriedYou won't get mad.  This'll be a fun one to solve.13:09
efriedCarry on.13:09
thorstfun eh?13:09
thorstsee 3894 and
thorstthat was / is a real mind bender.13:11
thorstand I hope I can just abandon them TBH13:11
*** kriskend_ has joined #openstack-powervm13:12
*** tjakobs has quit IRC13:13
*** seroyer has quit IRC13:25
*** tjakobs has joined #openstack-powervm13:33
*** tjakobs has quit IRC13:39
*** kotra03 has joined #openstack-powervm13:39
openstackgerritEric Fried proposed openstack/nova-powervm: HttpError=>VirtualInterface(Unp|P)lugException
*** seroyer has joined #openstack-powervm13:40
efriedthorst - whom should I add to ^^ to assess pvc impacts?13:40
thorstesha seth, eric larese, myself, svenkat13:41
*** kriskend_ has quit IRC13:45
thorstefried: I need to know something from you for hw13:54
thorstI have 7 SR-IOV cards left.13:54
efriedtell me13:54
thorstshould I do 4 in one system (super weird config...but...yeah), and then 1 card in 3 systems13:54
thorst2 cards in 3 systems and 1 in 1 system13:54
thorst2 cards in 2 systems and 3 cards in 1 system13:55
thorst1 card in 5 systems and 2 cards in 1 system <--- I'm leaning towards that13:55
efriedI would say 2/2/1/1/113:57
*** dwayne_ has joined #openstack-powervm13:57
efriedor 3/2/1/113:57 q13:57
thorstI'm going to go 2/2/1/1/1 (most systems - better for CI later)13:57
efriedBecause we want to be able to do LPM permutations that involve asymmetrical card setups13:58
thorstother many ports plugged in per card?  Both 10 gig ports?13:58
*** kriskend_ has joined #openstack-powervm13:58
efriedAll of them?13:59
thorstOK...that's kinda a pain13:59
efriedHow many ports on each card, two?13:59
thorsttwo I think13:59
thorstshould be fine...if they were 4 port I'd be grumpier13:59
efriedthorst, can we set 'em up one way to start, then move some around?14:01
thorstefried: yeah...but its a PITA14:01
thorstlab team justifiably gets frustrated when we do that14:01
thorstthe servers are heavy, the wiring is complex, etc...14:01
efriedSee, cause I think we should get some testing in with the following config:14:01
efried4 ports across 2-3 cards and 2 VIOSes.  Migrate back and forth between the 2- and 3-card system.  That oughtta work.14:01
thorstthey like one and done14:01
thorsthmm....I'd hope the PVC SVT is doing that.14:02
thorstwe can do that with 2 and 1 card systems?14:02
thorstcause each pport can go to one vios?14:02
efriedYeah, kinda, but that'd only be 2-way redundancy.  Different scenario.  Less depth exercising nvcastet's placement algo.14:03
efriedbbiab, mtg14:03
*** kairo has quit IRC14:05
*** tjakobs has joined #openstack-powervm14:10
*** apearson has quit IRC14:28
*** apearson has joined #openstack-powervm14:30
adreznecthorst: We're seeing "cloud-init-nonet[12.52]: waiting 120 seconds for network device" on deploys... that was related to the virt/interfaces.template ipv6 issue, right?14:30
thorstadreznec: yeppers14:32
adreznecHow did we fix that in the CI environment again?14:32
thorstwe have a parameter that we set to not do the RMC setting14:32
thorstso we lost RMC but got fast speed up14:33
*** esberglu has joined #openstack-powervm14:35
adreznecthorst: Parameter...14:36
thorstheh, sec14:37
adreznecSo... if we just don't create the MGMTVSWITCH14:37
thorstuse_rmc_ipv6_scheme = False14:37
adreznecWe wouldn't hit this, right?14:37
adreznecOr that14:37
adreznecWe need to get this fixed14:37
thorstset the parameter...14:37
thorstyeah, we do.  But how can we fix RHEL's cloud-init?14:38
thorstthis would be fine if the world just accepted IPv6  :-)14:38
thorstI wonder if you could put an Ipv4 floating IP on an IPv6 network.14:38
adreznecNot sure...14:38
thorstthe other fix is to just make your network IPv6.14:38
thorstat first I was kidding about that, but for what you're doing...14:39
thorstthat may actually be...desired?14:39
*** apearson_ has joined #openstack-powervm14:51
*** apearson has quit IRC14:53
*** apearson__ has joined #openstack-powervm14:56
*** apearson__ has quit IRC14:58
*** apearson_ has quit IRC14:58
*** apearson__ has joined #openstack-powervm15:00
*** apearson__ has quit IRC15:01
*** apearson__ has joined #openstack-powervm15:03
-openstackstatus- NOTICE: The Gerrit service on is restarting quickly to relieve resource pressure and restore normal performance15:22
*** k0da has quit IRC15:29
thorstefried: did you want to chat in here about the vif issue?16:12
efriedthorst, yes please.16:13
thorstso what'd you find out16:13
efriedWe have a CNA *and* a vNIC.16:13
efriedAre we supposed to have a CNA?16:13
efriedIt has a different MAC than the vNIC.16:13
efriedAnd that's the guy that's getting identified by PlugVifs.16:13
thorstahhhh ha!16:13
thorstOK - I got it16:13
thorstI understand the issue16:14
efriedSo the actual VIF shows up as "new" because... yeah.16:14
thorstopen a bug.16:14
thorstI'll work on it16:14
efriedI can nail it if you want.16:14
efriedI at least want to talk through the solution.16:14
thorsteither way16:14
thorstI'll let you16:14
efriedif vif['vnic_type'] == 'direct', get vNICs instead of CNAs?16:14
efriedIs that the right answer?16:14
thorstsec...pulling up code16:14
efriedNot sure what the return value is used for.16:14
thorstefried: so we always need get_cnas16:16
thorstbecause the gmt vifs16:16
thorstI think tasks/ line 123 (start of execute_impl)16:16
thorstwe could loop through the network_infos (which is just a list of vif objects)16:17
thorstand if one of those is type vNic, get the vnic_w_list...16:17
thorstto build the crt/update network_infos16:17
efriedthorst, 3900!16:21
efried"pretty" please16:21
thorstI assume there is some pun behind that16:22
*** esberglu has quit IRC16:31
efriedthorst, can we assume the PlugVifs Task is single-use?  Can I cache the CNAs/vNICs as instance vars?16:38
efriedSeems so, cause we're doing it for the update/crt network infos16:38
efriedjust want to double chec.16:38
*** apearson__ has quit IRC16:57
efriedthorst, also, the comment at the end of execute_impl implies that we're returning the CNAs we *created*.  In fact, we're only returning those that already existed before.  Is that correct?16:57
*** apearson__ has joined #openstack-powervm17:00
efriedthorst, it's used by PlugMgmtVif, which in fact probably needs *all* of the CNAs.  This looks like a regression to me.  Unless we're going to refactor all the plug methods to return the CNAs they're creating, the easiest thing to do would be to take vm_cnas out of PlugVif.provides and PlugMgmtVif.requires; and just have the latter do a full CNA.get17:04
thorstefried: Yes, single use.17:10
thorstefried: we should return all, and change wording17:10
efriedSeparate bug?17:10
thorstI'm OK with that, or just doing in same17:11
efriedthorst, Was planning to do it in same change set, since it's nigh inextricable; question was whether we need a separate launchpad bug (two Closes-Bug tags in same change set).17:13
thorstnah, just wrap it in17:13
thorstI'm ok with that17:13
efriedthorst, will this code path ever run where self.instance is of type VIOS?17:17
efried(ye gods, I hope not)17:17
thorstefried: no17:22
*** apearson__ has quit IRC17:26
*** tblakes has quit IRC17:28
openstackgerritEric Fried proposed openstack/nova-powervm: WIP: Create/update vNICs properly on Plug
efriedthorst ^^ -- pvc impacts?17:38
-openstackstatus- NOTICE: The infrastructure team is taking Gerrit offline for maintenance, beginning shortly after 18:00 UTC for a potentially 4 hour maintenance window.17:38
thorstefried: ^^17:58
efriedight.  Code is much cleaner now, IMO.17:59
efriedWell, "much" may be a stretch.  But cleaner.17:59
thorstwhich reminds I might not get to it17:59
thorstwe should remove all the import logging statements and switch over to oslo_log17:59
thorstthat may be a good one for kairo17:59
thorstI know he's been looking for some.17:59
efriedGosh, I thought we did that ages ago.18:00
thorstmissed a few places in nova_powervm18:00
thorstprobably specifically *I* missed a few places18:00
efriedonly four hits, yeah.18:00
thorstor reintroduced the bad bits18:00
*** apearson__ has joined #openstack-powervm18:02
-openstackstatus- NOTICE: Gerrit is now going offline for maintenance, reserving a maintenance window through 22:00 UTC.18:05
*** ChanServ changes topic to "Gerrit is now going offline for maintenance, reserving a maintenance window through 22:00 UTC."18:05
efriedthorst, PlugVifs.revert - shouldn't this only unplug the VIFs we plugged?18:23
thorstthis is sounding more and more like a bug18:23
adreznecGerrit's down, time to go home18:25
openstackLaunchpad bug 1619759 in nova-powervm "PlugVifs revert should only unplug what it plugged" [Undecided,New]18:31
efriedMay be getting a bit above my pay grade, but willing to tackle if you want to walk me through that second question.18:31
thorstefried: isn't that just changing the unplug to only iterate through the 'crt_vifs' list?18:32
efriedLook at the bug report.18:32
thorsto no...what're we getting at.18:33
thorsto nah...I don't think so.  The retreament only occurs on nova-compute start up18:34
*** k0da has joined #openstack-powervm18:37
*** kotra03 has quit IRC18:38
efriedthorst, my mistake - revert is doing the right thing.18:43
*** tblakes has joined #openstack-powervm18:44
*** ChanServ changes topic to "This channel is for PowerVM-related development and discussion. For general OpenStack support, please use #openstack."19:14
-openstackstatus- NOTICE: Gerrit is completing an online re-index, you may encounter slowness until it is complete19:14
*** miltonm has joined #openstack-powervm19:26
*** thorst has quit IRC19:36
*** thorst has joined #openstack-powervm19:37
*** thorst_ has joined #openstack-powervm19:39
*** thorst has quit IRC19:41
*** thorst_ has quit IRC19:43
*** kriskend_ has quit IRC19:55
*** seroyer has quit IRC19:57
*** seroyer has joined #openstack-powervm19:58
*** seroyer has quit IRC19:59
*** openstackgerrit has quit IRC20:04
*** openstackgerrit has joined #openstack-powervm20:04
*** tjakobs has quit IRC20:05
*** seroyer has joined #openstack-powervm20:14
*** tblakes has quit IRC20:32
openstackgerritEric Fried proposed openstack/nova-powervm: Create/update vNICs properly on Plug
*** thorst has joined #openstack-powervm20:46
*** dwayne_ has quit IRC20:52
efriedthorst, yrt?20:53
thorstwhat's up20:53
efriedpossible for self.network_infos to be None?20:53
thorstwell, maybe not none, but empty20:53
efriedk, adding harmless safety.20:59
efriedthorst, what's your availability going to be over the next few business days wrt code reviews?21:00
thorstI'll be on a plane for most of the next few days21:01
efriedk. is relatively urgent; the other not so much.21:02
efrieddidn't we implement something in power-off that wouldn't dump the huge glut of pypowervm cruft if the partition doesn't exist?21:03
thorstefried: I think it was something that esberglu was proposing to nova-powervm21:04
*** tjakobs has joined #openstack-powervm21:06
efriedthorst, fyi, on .149, vnic create is taking 16-19s; delete 10-11s.21:08
thorstthat's not great....but not the worst thing in the world21:08
*** tjakobs has quit IRC21:13
efriedthorst, finally got done with those two reviews you gave me this morning (on the migration slot mappings for NPIV)21:32
thorstefried: saw...and yeah, they need work21:32
thorstor the core has a change it could make too21:33
thorstboth options are..meh21:33
efriedOkay.  Well, it took me all day, but I think I get what you're doing here.21:33
efriedStill some pending questions of design, though I think reviewing the community one answered my boggles in the pypowervm one.21:34
efriedClearly you're expecting the same number of VIOSes at the destination - was that always the expectation?21:34
thorstefried: it is now21:34
thorstnot really just that, but that the same slots line up on a target VIOS21:36
thorstefried: 365078...21:38
thorstare you assuming that all the network_info are direct or sea?21:38 doesn't look like you are21:39
efriedI'm assuming anything that is 'direct' needs to look for VNIC; and anything else needs to look for CNA.21:39
thorstthe _vif_exists confused me21:39
efriedDid I misname that?21:39
thorstno, you're fine...I just confused myself21:39
efriedI was using 'vif' as an umbrella for "CNA or VNIC (or possible future other thingy)"21:39
*** k0da has quit IRC21:44
thorstefried: just one thing I don't like21:44
efriedJust one?  I did well.21:44
efriedTell me.21:44
thorstyou made the pipeline slower21:45
thorstI know why you did this21:45
thorstbut I disagree it has to be done21:45
efriedYou mean because I'm doing a fresh CNA feed get in PlugMgmt?21:45
efriedThe alternative was to accumulate all the CNAs as I create them.21:46
efriedWhich would entail changing the spec of alllll the plug methods to return said CNAs.21:46
efriedWhich I could do.21:46
efriedBut a CNA feed get is really fast, isn't it?21:46
efriedhence not worth the extra code & risk therein entailed21:47
efriedThat was my logic, anyway.21:47
thorstefried: to a VM...yeah, it can be21:49
thorstbut why bother?21:49
thorstits like 3 lines less code21:50
thorstbut a feed get otherwise21:50
efriedsorry, you lost me.21:50
thorstits like half a second...21:50
thorstso I'd rather just write the three loc21:50
thorstback in a bit21:50
*** tjakobs has joined #openstack-powervm21:55
efriedthorst, okay, I see, the leaf plug methods are already returning the vifs - it's just nova_powervm.virt.powervm.vif.plug that would need to return it back up.21:56
*** tjakobs has quit IRC21:59
*** apearson__ has quit IRC22:14
*** tjakobs has joined #openstack-powervm22:16
*** tjakobs has quit IRC22:21
*** miltonm has quit IRC22:31
openstackgerritEric Fried proposed openstack/nova-powervm: Create/update vNICs properly on Plug
efriedthorst ^^ (more than 3LOC, btw)23:11
efriedtested, works.23:16
efried(though I didn't test it through all possible permutations, by any means)23:16
efriedSigning off.  Feel free to merge it if you're happy.23:17
*** openstackgerrit has quit IRC23:18
*** openstackgerrit has joined #openstack-powervm23:19
thorstefried: LGTM23:28
*** thorst has quit IRC23:30
*** thorst has joined #openstack-powervm23:32

Generated by 2.14.0 by Marius Gedminas - find it at!