Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-06-12
16:06:24 efried sahid: Yeah, figured that wasn't enough to hold it up, I would fast-approve a fup to fix that if you're so inclined.
16:08:19 sahid efried: i'm not
16:08:24 efried :)
16:08:33 openstackgerrit Matt Riedemann proposed openstack/nova master: WIP: Add method to get cpu traits https://review.openstack.org/560317
16:09:32 cfriesen mriedem: sorry, you referring to the swap size thing?
16:09:44 mriedem cfriesen: no, the "wtf test failure" thing
16:10:05 mriedem oh oops
16:10:07 mriedem cdent: ....
16:10:10 mriedem sorry, too many c's
16:10:41 mriedem cfriesen: for the ceph swap size thing, i'd ask mnaser
16:10:50 mnaser hi
16:10:50 cfriesen no worries, I thought I had missed some context. :)
16:11:42 openstackgerrit Brianna Poulos proposed openstack/nova master: Implement certificate_utils https://review.openstack.org/479949
16:11:44 openstackgerrit Brianna Poulos proposed openstack/nova master: Plumb trusted_certs through libvirt driver image paths https://review.openstack.org/561262
16:11:44 openstackgerrit Brianna Poulos proposed openstack/nova master: Add trusted_image_certificates to REST API https://review.openstack.org/486204
16:11:45 openstackgerrit Brianna Poulos proposed openstack/nova master: Add notification support for trusted_certs https://review.openstack.org/563269
16:11:46 openstackgerrit Brianna Poulos proposed openstack/nova master: Add certificate validation docs https://review.openstack.org/560158
16:11:49 mnaser cfriesen, mriedem: we don't use swap (swap on ceph sounds rough) but do you have a bit more context or bug #?
16:12:19 mriedem mnaser:
16:12:20 mriedem (10:15:56 AM) cfriesen: is anyone aware of an issue where an RBD-backed swap disk is not restored to the initial size after a resize/revert? We're hitting this on Pike.
16:12:20 mriedem (10:16:47 AM) cfriesen: also, we're seeing a second issue where the guest has to manually re-run "mkswap" after a resize/confirm otherwise it still keeps using the old size of swap
16:12:36 cfriesen mnaser: haven't opened a bug yet, wanted to see if it was a known issue
16:12:54 mriedem mnaser: you use the rbd imagebackend don't you for ephemeral local disks?
16:12:58 mnaser mriedem: yes
16:13:05 mriedem how do you prevent users from attaching swap devices?
16:13:18 mnaser well, they could do it themselves by adding a volume
16:13:32 mriedem right, it's part of the block_device_mapping_v2 request when creating a server
16:13:32 mnaser i thought we're talking about the swap flavor attribute (that's a thing.. right?)
16:13:40 mriedem it's related
16:13:55 mriedem the swap bdm gets it's size from the swap value in the flavor
16:14:31 sahid efried: actually it would have been better to use a constant instead of 0, also since we have changed the method name it would have been better to refer a name with xen
16:14:58 cfriesen in the second case, I'm wondering whether we're copying/expanding the original ceph volume and not re-running mkswap on it, so the guest's on-disk data structures still show the "old" size
16:15:10 efried sahid: I can pull it out of the gate if you like.
16:15:41 mriedem https://github.com/openstack/nova/blob/master/nova/compute/api.py#L1368
16:15:47 mnaser i think resize doesn't touch the underlying devices
16:15:50 mriedem i think you can create swap bdms with their own size not from the flavor
16:15:57 mriedem https://github.com/openstack/nova/blob/master/nova/compute/api.py#L1375
16:16:06 mnaser same reason how resizing will result in a drive larger than the partiton
16:16:28 mriedem but if the flavor doesn't have swap set, then i guess that will always fail
16:16:29 sahid efried: what do you think? there are enough nits to hold the merge?
16:16:45 efried sahid: Would you have put a -1 on it for these issues?
16:17:14 dansmith efried: sahid I just commented on both
16:17:20 dansmith unsurprisingly, I think it's fine as-is
16:17:47 dansmith you're going to reset half the gate queue if you bump it out, FYI
16:18:09 sahid yes probably, for the constant and the test introduced that use nested
16:18:35 sahid but yes if the gate is already running i don't think any problem :)
16:18:53 zcorneli sahid: I can fix these as well within the next patch in the series.
16:18:53 sahid s/think/see
16:19:24 dansmith zcorneli: don't pollute your next patch with cleanups, do a separate patch from it
16:19:28 sahid zcorneli: that could be unrelated, but in a nother patch yes
16:19:32 cfriesen mnaser: isn't swap different since it can be wiped out during the resize and the guest is fine with that?
16:20:01 mnaser cfriesen: i'm assuming the code base doesn't actually touch anything if it's a resize
16:20:08 zcorneli dansmith: Ok, no worries
16:20:16 mnaser cfriesen: libvirt i assume?
16:22:28 cfriesen mnaser: yep
16:23:02 openstackgerrit Lee Yarwood proposed openstack/nova master: Add regression test for bug #1764883 https://review.openstack.org/562072
16:23:03 openstack bug 1764883 in OpenStack Compute (nova) "Evacuation fails if the source host returns while the migration is still in progress" [Undecided,In progress] https://launchpad.net/bugs/1764883 - Assigned to Lee Yarwood (lyarwood)
16:23:03 openstackgerrit Lee Yarwood proposed openstack/nova master: compute: Ensure pre-migrating instances are destroyed during init_host https://review.openstack.org/562284
16:23:09 mriedem is the libvirt driver the only one (really) that supports swap disks? because that's what the api-ref says:
16:23:10 mriedem "Specifies the guest server disk file system format, such as ext2, ext3, ext4, xfs or swap. This parameter affects only the libvirt virt driver."
16:25:20 mriedem maybe that just means the non-swap values are only used by the libvirt driver
16:25:40 dansmith or that libvirt is the only one that makes filesystems on disks
16:25:46 mnaser ^
16:25:58 dansmith does't mean you don't get a swap disk according to the flavor,
16:26:08 dansmith you just might have some-assembly-required sort of deal
16:26:35 mriedem ok, it
16:26:50 mriedem it's just that "This parameter affects only the libvirt virt driver." is confusing since you have to specify guest_format='swap' to get a swap bdm attached
16:26:59 mriedem along with some other bdm field incantations
16:27:04 mriedem that aren't documented in the api ref
16:32:41 mnaser https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L8180-L8186
16:33:21 mnaser https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L8235-L8241
16:33:38 mnaser so because it already exists with ceph, it probably doesnt touch it
16:34:12 mnaser (i think..)
16:34:55 tasker hey, good morning. I'm checking up on https://review.openstack.org/#/c/550865/ . is there something else needed before it can be merged? also, is this review a candidate for backport to Pike?
16:34:59 cfriesen mnaser: we're seeing the underlying size change to the "new" size when doing a resize...but the guest needs to re-run "mkswap" to make use of the larger amount of space.
16:36:08 cfriesen mnaser: and if we revert the resize then the underlying size stays at the "new" size even after reverting, which would seem to be a resource tracking bug.
16:37:27 mnaser i can't even find the codebase where nova does mkfs
16:37:31 mnaser i can find the function calls..
16:38:52 mnaser ok _create_swap()
16:39:15 mnaser looks like mkswap only happens in _create_swap and nowhere else
16:39:27 mnaser idk if we want to call that a bug or not
16:39:42 mnaser i'd want to avoid touching underlying fs once something is booted
16:45:26 mriedem jaypipes: i made that server_default=text("0") model change for a reason, to match the db schema migration script; the fact it passes with sqlite and postgresql makes me think it's a bug in whatever mysql comparison code is happening, which zzzeek seems to agree with
16:45:44 tasker https://review.openstack.org/#/dashboard/self is only showing open reviews. isn't there a way to show all reviews that I've submitted?
16:46:21 mriedem tasker: https://review.openstack.org/#/q/owner:eric%2540awnix.com ?
16:46:41 tasker oh.
16:46:44 tasker thanks!
16:51:35 openstackgerrit Matt Riedemann proposed openstack/nova master: api-ref: expand on various bdm parameters https://review.openstack.org/574805
16:53:20 jaypipes mriedem: k. I haven't pushed any further update on that series...
16:53:44 mriedem jaypipes: having said that, i'm not sure what to do here, since i don't want to hold things up while a bug in a dependent library gets fixed
16:53:58 openstackgerrit Simon Dodsley proposed openstack/nova master: Add enhanced KVM storage QoS quotas https://review.openstack.org/558530
16:53:59 mriedem i guess i can respin, remove the model change, add a FIXME and report a bug to oslo.db
16:54:03 jaypipes mriedem: maybe we just put a TODO in there?
16:54:13 mriedem yeah i'll do that quick
16:54:15 jaypipes mriedem: or I can do it. just let me know.
16:54:31 mriedem i've got it local, 2 seconds
16:54:41 jaypipes k, thx mriedem
16:55:55 simondodsley_ mriedem: cinder docs review added https://review.openstack.org/#/c/574804/
16:55:56 cfriesen mnaser: for what it's worth, for "local" filesystems we don't copy over the swap file but instead create a whole new one, so the guest sees the "right" swap size when it boots up after the resize.
16:56:30 mnaser cfriesen: ah you are right about that one, so we do kinda make the drive disappear with resize

Earlier   Later