Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-05-23
16:06:07 mriedem yes
16:06:29 mriedem well i know they are neutron,
16:06:34 mriedem i assume they have a single port on the instance
16:06:40 mriedem auto-created by nova when creating the server
16:08:03 cfriesen is the test using get-me-a-network? or could it be booting with no port?
16:09:33 mriedem no it's not those
16:11:15 cfriesen mriedem: on of our neutron guys says it might happen if nova didn't update the host_id, or if there was an issue with vif-unplug/vif-plug on either side.
16:14:40 mriedem if that happened, i'd think the job would fail
16:14:44 mriedem or we'd see errors in the logs
16:14:56 mriedem granted, these tests don't actually try to ssh into the guest once it's migrated to the target host
16:20:44 openstackgerrit Stephen Finucane proposed openstack/nova master: console: introduce basic framework for security proxying https://review.openstack.org/345396
16:20:45 openstackgerrit Stephen Finucane proposed openstack/nova master: console: introduce the VeNCrypt RFB authentication scheme https://review.openstack.org/345398
16:20:45 openstackgerrit Stephen Finucane proposed openstack/nova master: console: introduce framework for RFB authentication https://review.openstack.org/345397
16:20:46 openstackgerrit Stephen Finucane proposed openstack/nova master: console: provide an RFB security proxy implementation https://review.openstack.org/345399
16:24:28 openstackgerrit Matt Riedemann proposed openstack/nova master: doc: start a FAQs section for cells v2 https://review.openstack.org/467309
16:24:31 mriedem kfox1111: ^
16:24:34 mriedem i said i'd do it
16:28:31 edleafe jaypipes: regarding your comment on UUID passing in https://review.openstack.org/#/c/465171/, that was my original approach. However, dansmith preferred to see it as an additional RPC param
16:29:02 jaypipes edleafe: well, based on this ML thread, we may have to rethink this all anyway...
16:30:32 edleafe jaypipes: not sure I follow (I'm still catching up from PyCon)
16:30:59 edleafe jaypipes: We can claim in the scheduler to make sure we have at least one sufficient host
16:31:24 edleafe jaypipes: and then return that host and some 'alternates'
16:31:52 edleafe that would get passed down to the cell, which could retry an alternate in the event of a local build failure
16:32:09 edleafe Where is the problem under discussion?
16:32:21 jaypipes edleafe: I'm referring to this thread http://lists.openstack.org/pipermail/openstack-dev/2017-May/117221.html
16:32:51 jaypipes edleafe: http://lists.openstack.org/pipermail/openstack-operators/2017-May/013464.html
16:33:05 edleafe jaypipes: yes, I've read all that thread
16:33:06 jaypipes edleafe: I'd like to get rid of retries entirely.
16:33:08 openstackgerrit Dan Smith proposed openstack/nova master: Fix live migration devstack hook for multicell environment https://review.openstack.org/458537
16:33:08 openstackgerrit Dan Smith proposed openstack/nova master: De-complicate some of the instance delete path https://review.openstack.org/458636
16:33:09 openstackgerrit Dan Smith proposed openstack/nova master: Target cell on local delete https://review.openstack.org/464059
16:33:23 jaypipes edleafe: but ops are crapping on the idea.
16:33:32 openstackgerrit Eric Berglund proposed openstack/nova master: Add strict option to discover_hosts https://review.openstack.org/467218
16:33:56 edleafe jaypipes: so wouldn't adding a list of alternates be an acceptable compromist?
16:34:04 dansmith edleafe: that was the compromise
16:34:14 dansmith edleafe: jaypipes was apparently missing that bit of context from the summit
16:34:16 mriedem you know, if you're providing a monthly subscription service, you'd think you'd want your website to work such that your customers can update their credit card information to actually pay you, rather than cancel their subscription, even after you've submitted a ticket and called customer service
16:34:46 jaypipes edleafe: I would still like to do retries (from resource allocation failures due to concurrent claims) in the scheduler.
16:35:06 edleafe jaypipes: yes, that's what I've done in my series
16:35:09 dansmith jaypipes: you shouldn't call those retries, IMHO, to avoid confusion
16:35:11 jaypipes edleafe: I know :)
16:35:17 jaypipes dansmith: touche
16:35:21 edleafe so what is the problem?
16:35:56 jaypipes edleafe: I'm still having a tantrum about not getting my way with ops.
16:35:57 edleafe dansmith: I like to call them 'claims' :)
16:36:06 jaypipes edleafe: gimme ten minutes to come to my senses please.
16:36:24 edleafe jaypipes: ok. Grab your blankie and open a juice pack
16:36:41 jaypipes thanks guys.
16:39:48 openstackgerrit Matt Riedemann proposed openstack/nova master: doc: add cells v2 FAQ on refreshing global cells cache https://review.openstack.org/467318
16:46:24 openstack Launchpad bug 1692984 in neutron "VM boot fails with "TCG doesn't support requested feature: CPUID.01H:ECX.vmx"" [High,New]
16:46:24 ihrachys anyone seen the failure? https://bugs.launchpad.net/neutron/+bug/1692984
16:46:58 mriedem ihrachys: nope, libvirt/qemu fun
16:47:00 mriedem kashyap: ^
16:47:18 ihrachys I am not completely sure this is the culprit. the log suggests it's a 'warning'
16:47:35 ihrachys so *maybe* it just spills it as part of stderr when some other failure happens.
16:47:45 ihrachys I am too dumb to distinguish :)
16:49:42 mriedem please standby, firefox is struggling to open logs
16:50:31 mriedem ihrachys: it definitely killed the server build http://logs.openstack.org/83/466983/1/check/gate-tempest-dsvm-neutron-linuxbridge-ubuntu-xenial/bd69e6b/logs/screen-n-cpu.txt.gz?level=TRACE#_May_23_03_04_29_688986
16:51:55 ihrachys yeah, sure. I am merely saying I am not sure if the cpu flag thing is a red herring, or the culprit of the failure
16:52:35 jaypipes cdent: in a gabbit, can I test for greater or equal to a value?
16:52:55 jaypipes cdent: for example, $.traits.`len`: 2
16:53:07 jaypipes cdent: can I change that to test for >= 2?
16:53:14 cdent maybe, I need to do some checking
16:53:34 mriedem ihrachys: hmm, yeah http://logstash.openstack.org/#dashboard/file/logstash.json?query=message%3A%5C%22process%20exited%20while%20connecting%20to%20monitor%3A%20warning%3A%20TCG%20doesn't%20support%20requested%20feature%3A%20CPUID.01H%3AECX.vmx%20%5Bbit%205%5D%5C%22%20AND%20tags%3A%5C%22screen-n-cpu.txt%5C%22&from=7d
16:53:41 mriedem ihrachys: it's definitely not 100% failure
16:53:43 mriedem not even close
16:56:19 ihrachys right. so I was thinking, maybe it's just spilled all the time, but we dump stderr on failures. but other log messages are very generic to my taste, and I needed to capture something in the bug, so...
16:57:24 cdent jaypipes: simple answer is "not really" but you could perhaps do a regex to get the results you want?: /[2-9]/
16:57:58 openstackgerrit Matt Riedemann proposed openstack/nova master: doc: add cells v2 FAQ on mapping instances https://review.openstack.org/467322
16:58:41 cdent jaypipes: the limitation at this point is that the RHS is that the assert is either an assertEqual or an assertRegex
16:58:41 mriedem pabelanger: dansmith: fyi, no one replied to my question about doing PUT for keypairs in the ops channel
17:12:16 jaypipes cdent: k, no prob. thx!
17:21:53 jaypipes mriedem, smcginnis: if it makes you feel any better (doubtful), k8s is dealing with similar feature requests and inconsistent UX around volume types. There, it's called dynamicStorageClass but the concept is virtually identical
17:22:52 cfriesen jaypipes: got a sched race for you to consider for placement/claims: 2 vms in server group with anti-affinity, both vms get migrated at same time, scheduler picks same host for both, both end up on same host.
17:23:23 jaypipes cfriesen: two trains leave New York and San Francisco, each going 60mph...
17:23:27 cfriesen jaypipes: make that live-migrated. ^
17:24:10 cfriesen not sure how to fix this without either atomic schedule/claim or secondary check on compute and possible reschedule
17:24:34 cfriesen (I guess secondary check could be anywhere post-claim)
17:25:00 jaypipes cfriesen: we can dramatically decrease the length of time that such a decision can be made with claims in the scheduler...
17:26:18 cfriesen jaypipes: agreed,
17:27:49 cfriesen but it might still make sense to allow reschedule within the cell
17:30:45 jaypipes cfriesen: yes, perhaps.
17:31:00 jaypipes cfriesen: trying to avoid that though if possible. (doesn't look like we'll be able to, though)
17:33:53 pabelanger mriedem: dansmith: Ya, for now we are leaning towards updating ansible os_keypair module to support the update functionality
17:34:10 pabelanger then delete / create the key with same name
17:40:45 cfriesen jaypipes: something to consider in the above scenario...if we have an anti-affinity instance migrating from host1 to host2, we probably want to exclude *both* hosts from consideration when scheduling other instances in the same group.
17:42:27 jaypipes cfriesen: sure, no disagreement.
18:00:40 ildikov lyarwood: sfinucan: hi
18:00:46 ildikov lyarwood: sfinucan: if you have some time there're two patches related to the new Cinder attach/detach flow with one +2 already: https://review.openstack.org/#/c/459111/ https://review.openstack.org/#/c/459113/
18:00:55 ildikov lyarwood: sfinucan: tnx :)
18:06:48 openstackgerrit OpenStack Proposal Bot proposed openstack/nova master: Updated from global requirements https://review.openstack.org/467344
18:10:26 efried jaypipes johnthetubaguy mordred Would y'all please have a look at this spec update? Been +2 for a few weeks. https://review.openstack.org/#/c/461481/
18:12:02 efried But the pen IS mightier than the sword, so...
18:12:07 jaypipes :)
18:14:03 jaypipes efried: done.
18:14:10 efried jaypipes Thanks dude.
18:16:42 efried jaypipes Still on my list to ask you for that pointer to the shared resource pool code. Is that a quick answer?
18:17:17 jaypipes efried: https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/shared-resources-pike
18:20:58 efried jaypipes Okay, looking through. I know you've slept (in multiple time zones) since then, so to refresh your memory on the conversation: I'm looking to get my compute driver itself to do the registration of its shared storage pool.
18:21:21 jaypipes efried: ah. yeah, sorry.. one sec

Earlier   Later