Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-05-31
21:33:51 edleafe melwitt: looking now...
21:34:17 melwitt thanks edleafe
21:35:29 melwitt dansmith: looking at the live migration hook thing, what was failing and how do we see it fixed by https://review.openstack.org/#/c/458537 ?
21:35:39 openstackgerrit Merged openstack/nova master: PowerVM Driver: SSP emphemeral disk support https://review.openstack.org/443189
21:35:55 dansmith melwitt: you only see it when you run with the devstack patch.. that one had a depends-on that devstack patch until the latest rev,
21:36:06 dansmith so you can look at some of the early patches,
21:36:18 dansmith or we can do a dummy patch that depends on the devstack patch and not this one if you want to see it
21:36:25 melwitt oh, I see. I'll do that, just haven't been in the loop of what is going on there
21:36:29 dansmith but this has to go before that one, so I broke the link now that it's clear it's good
21:36:31 dansmith yeah
21:36:35 dansmith s'all good
21:38:50 mriedem dansmith: heh, and since then,
21:38:58 mriedem we've disabled running ceph in the live migration job anyway
21:39:05 dansmith ah
21:39:14 mriedem https://github.com/openstack/nova/commit/994caa464efdbed3d684354d699cd79e41d7eb0f
21:39:18 dansmith well, this is correctness of devstack stuff, so probably still useful, but whatever you want to do
21:39:20 mriedem oh wait,
21:39:21 mriedem only for grenade
21:39:38 mriedem still runs for master live migration job non-grenade
21:42:30 mriedem dansmith: shouldn't NOVA_CPU_CONF default to NOVA_CONF and then default to the hard-coded path?
21:42:32 mriedem to be compatible?
21:42:56 mriedem ${NOVA_CPU_CONF:-${NOVA_CONF:-/etc/nova/nova.conf}}
21:42:59 mriedem or something
21:43:33 mriedem chances of that being a real world problem are slim to none though
21:44:15 dansmith not as of my patch in devstack
21:44:21 dansmith unless you mean before the patch in devstack lands?
21:44:30 mriedem correct
21:44:44 dansmith Hmm, I dunno I guess
21:44:50 dansmith I guess we wait and see if it fails?
21:46:46 mriedem cdent: passed http://logs.openstack.org/36/469636/3/check/gate-swift-dsvm-functional-ubuntu-xenial-nv/2f34544/logs/devstacklog.txt.gz#_2017-05-31_21_14_48_452
21:47:44 melwitt mriedem: which log file in the live migration job shows where ceph runs? I can never remember
21:47:56 mriedem console
21:48:14 melwitt ah, there we go. thanks
21:48:34 melwitt I missed it the first time I looked, somehow
21:49:26 mriedem jaypipes: +W on https://review.openstack.org/#/c/469625/
21:49:32 mriedem we'll get a release out once that's merged
21:56:51 openstackgerrit Matt Riedemann proposed openstack/nova master: Skip micorversion discovery check for update/delete volume attachments https://review.openstack.org/469668
21:57:11 openstackgerrit Matt Riedemann proposed openstack/nova master: Skip microversion discovery check for update/delete volume attachments https://review.openstack.org/469668
21:57:18 dansmith eff
21:57:20 dansmith gdi
21:57:23 dansmith so wanted to -1 mriedem for a typo
21:57:24 mriedem ha
21:57:40 mriedem you've got to be fast
21:57:52 dansmith 20 seconds seemed pretty fast
21:58:34 cfriesen speaking of typos, how's his test covfefe?
21:59:39 mriedem it's excellent
21:59:47 mriedem fabulous
21:59:48 mriedem the best
22:02:23 mriedem jaypipes: here you go https://review.openstack.org/469670
22:04:53 jaypipes mriedem: danke
22:24:18 mriedem efried: does https://review.openstack.org/#/c/443189/ mean we can close that bp for pike?
22:24:23 mriedem or is it carte blanche?
22:25:11 efried mriedem I wouldn't mind getting more function in; does closing the bp preclude that?
22:25:57 efried mriedem Also, there will be some small followup work on the already-merged codebase (from code comments in previous changes, following community stuff like resource provider work, etc.) - assume that stuff just gets to be regular non-bp-linked change sets.
22:27:01 mriedem the follow up cleanups are fine to be separate from the bp,
22:27:18 mriedem and i think the get_inventory() stuff can be tracked under whatever was used for the same thing in the libvirt driver
22:29:45 efried mriedem So if we want to propose more function, and the bp is closed, will those changes automatically get -2ed and punted to queens? (By "automatically" I of course refer to the turd in your pocket.)
22:32:58 mriedem i wouldn't close it then, but i'd rather it not churn on forever either
22:33:18 mriedem i want to close things out so we can focus reviews on stuff that's not done yet
22:40:26 efried mriedem Okay, let me go back to the team and see if we want to shoot for more function in pike (config drive is small; and OVS would give us an actual usable VM). Tomorrow soon enough for me to get back to you on this?
22:40:57 mriedem efried: yeah i'm not in a rush
22:41:01 efried k
22:45:57 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Default AZ for instance if cross_az_attach=False and checking from API https://review.openstack.org/469675
22:48:02 mriedem god cross_az_attach is terrible
22:48:07 mriedem woe to anyone that ever sets it to False
22:50:51 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Default AZ for instance if cross_az_attach=False and checking from API https://review.openstack.org/469675
23:24:11 tasker I'm having a problem with freezing the filesystem using the libvirt modules. I keep getting "AttributeError: 'virDomain' has no attribute 'fsFreeze'". I'm using Nova Mitaka (13.1.2), and a full grep for 'fsFreeze' only shows the failing call at line 440, the definition in the mocklibvirt driver for the test, and a couple of comments. The libvirt driver itself (1.3.2) has no definition for a
23:24:17 tasker function named 'fsFreeze'. The only bug report I've found referencing is from last May and says to use libvirt >= 1.2.5.. what am I missing? anyone have any ideas?
23:27:07 clarkb tasker: what version of libvirt and what version of libvirt-python?
23:27:48 tasker using OSA, the venv has 1.3.2, the dpkg for the system wide is 1.3.1
23:28:18 mriedem you need at least libvirt 1.2.5
23:28:35 cfriesen_ can you run "virsh domfsfreeze" manually on the compute node?
23:28:39 mriedem fsFreeze is a function on libvirt-python i believe
23:28:40 tasker nova calls oslo_utils.importutils.import_module("libvirt") which pulls in the 1.3.2 found in /openstack/venvs/nova/lib/python2.7/site-packages/libvirt.py
23:29:04 cfriesen_ what version is your actual libvirt though?
23:29:14 mriedem you also need hw_qemu_guest_agent in the image
23:29:52 cfriesen_ tasker: what does "virsh --version" give?
23:30:13 tasker 'virsh --version' reports "1.3.1"
23:30:25 tasker and I can call and issue domfsfreeze manually.
23:30:55 mriedem seems like a problem with the libvirt-python package,
23:31:03 tasker myproblem is that there is no function called fsFreeze.
23:31:19 mriedem people have reported issues like this before where libvirt and the libvirt-python package say they are at a certain version but the python binding is actually old or missing some things
23:31:34 clarkb the way libvirt-python works is it exposes the features of libvirt that it was built against
23:31:36 mriedem i'm not really sure how
23:31:36 cfriesen_ tasker: what's your "virt_type"?
23:31:48 clarkb so if you installed via a wheel or similar that was built against an old version of libvirt that could explain it
23:31:54 tasker virt_type = 'kvm'
23:32:12 tasker clarkb: i like that explanation. it makes a lot of sense.
23:32:29 clarkb pip install --no-binary -U libvirt-python should force it to rebuild though that will affect whatever system you build against
23:32:36 mriedem https://bugs.launchpad.net/ubuntu/+source/nova/+bug/1539506
23:32:36 openstack Launchpad bug 1539506 in Ubuntu Cloud Archive liberty "AttributeError: 'virDomain' object has no attribute 'fsFreeze" [Medium,Fix committed]
23:33:25 tasker mriedem: that bug only says what you've mentioned regarding libvirt versions and where to get updated ones. I did not find it helpful.
23:33:42 mriedem ok i can't tell if they put out an updated libvirt-python for mitaka or not
23:33:50 tasker your's and clark's mention of misbuilt packages makes a lot more sense.
23:34:06 tasker neither could I; hence my coming here.
23:34:29 tasker clarkb: I'll give that pip incantation a shot.
23:35:24 tasker could you explain how doing that would affect "whatever system you bulid against" in some more detail?
23:35:58 clarkb tasker: it will overwrite any existing libvirt-python install potentially breaking other working functionality :)
23:36:04 clarkb It should be safe though
23:36:17 tasker ahh .. ok. so backups for everyone!

Earlier   Later