Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-07-08
15:42:53 mriedem note that _move_claim will have the same issue here i'd think
15:43:07 cdent a) what mriedem said, b) the idea is to make sure placement has a way to prevent someone asking for more physical disk than is available
15:43:45 cdent mriedem: it's less of a risk there as they a) don't happen as much, b) happen even less in the vmware environment, c) this problem really only shows up when throwing > 2000 vms as the same nova-compute
15:43:57 cdent s/as the/at the/
15:44:11 sean-k-mooney cdent: ya i get that. i wonder if this is somthing we coudl do when we make the inital allocation
15:44:29 mriedem the initial allocation doesn't involve the driver
15:44:30 sean-k-mooney e.g. is this somethign we shoudl consider for all driver in general
15:44:37 sean-k-mooney true
15:44:41 efried kashyap: Thank you. Your votes/comments on those patches will be beneficial. Sounded like melwitt will also appreciate it.
15:44:53 cdent no other driver (that I'm aware of) would need this as they report real and true inventory
15:44:56 sean-k-mooney but shoudl we be doing it for libvirt with ceph/nfs
15:45:21 cdent it's only when you have non-contiguous disk(s)
15:45:52 sean-k-mooney cdent: ok would this be solved in the futrue by representing
15:45:54 mriedem in the libvirt+ceph case, we report total and max_unit the same, and placement will reject requests that are too big because of the capacity calculation in placement, right?
15:46:00 sean-k-mooney the datastores as different shareing providers
15:46:27 cdent sean-k-mooney: yes, except for the thing I said above about something called "storage policy" which would disrupt placement's understanding of where things are
15:46:36 sean-k-mooney mriedem: yes which is fine for ceph as we have a singel pool for vms
15:47:02 mriedem i'd prefer not to have a config option for this behavior,
15:47:07 mriedem you could add a variable on the driver itself,
15:47:22 mriedem which could be controlled by the driver depending on if the vcenter driver is doing this shared storage pool modeling thing
15:47:24 cdent yeah, I'm not even clear if it should ever be upstreamed
15:47:30 mriedem the RT would check that, and by default we do as we do today
15:47:49 mriedem i've done similar hacks in the RT for ironic
15:47:54 mriedem b/c they are a special unicorn as well
15:48:02 mriedem *we've done
15:48:21 cdent unicorn's are a PITA
15:48:28 cdent hmm unicorns too
15:48:48 mriedem always pooping out frosting on my yard and such
15:48:59 cdent Thanks mriedem, sean-k-mooney, efried this has been very useful.
15:49:07 cdent and such
15:49:15 kashyap efried: Yes, will review
15:49:43 cdent If it turns out I can extract a little bit of 'if this virtdriver magic' i'll make it so
15:50:41 openstackgerrit Sundar Nadathur proposed openstack/nova master: ksa auth conf and client for cyborg access https://review.opendev.org/631242
15:50:42 openstackgerrit Sundar Nadathur proposed openstack/nova master: WIP: Create and bind Cyborg ARQs. https://review.opendev.org/631244
15:50:42 openstackgerrit Sundar Nadathur proposed openstack/nova master: WIP: Add Cyborg device profile groups to request spec. https://review.opendev.org/631243
15:50:43 openstackgerrit Sundar Nadathur proposed openstack/nova master: WIP: Get resolved Cyborg ARQs and add PCI BDFs to VM's domain XML. https://review.opendev.org/631245
15:53:44 mriedem gibi: looks like https://review.opendev.org/#/c/637955/31 and https://review.opendev.org/#/c/669188/2 conflict, maybe you want to rebase the fix underneath the bigger change?
15:57:09 sean-k-mooney cdent: by the way i assume there are buisness reasons that vmware would prefer not to map each compute node to a specifc data store and use cinder with volume type to express storage policies for everything else
15:57:39 cdent sean-k-mooney: yeah, the basic root of all these things is "let the DRS continue to work"
15:57:47 sean-k-mooney i know vmware have lots of fancy feature built into the hypervior and storage solution so im guessign the current approch is trying to expose them
15:58:02 sean-k-mooney right ok makes sense
15:58:31 cdent it creates a huge impedance mismatch between the way nova thinks and the way vmware thinks, but that's the way it goes
15:59:48 sean-k-mooney ya hyperv also has some limited clustering of vms and storage pools but its not as advanced as what vcenter does so the impedence miss match is less of an issue for them i guess
16:00:43 cdent there's been discussion of tying to listen to events from the DRS via some agent that would then "correct" placement when DRS makes a change
16:01:14 cdent but that would require more effort than currently available resources
16:01:45 gibi mriedem: will do
16:01:55 sean-k-mooney that might require you to reshap allocation if it moved things right if you had a shareing provider per datastore
16:02:30 sean-k-mooney although if you had just one RP and hid the datastore i guess it woul just have to modify the max unit
16:02:35 sean-k-mooney as it does today
16:03:12 sean-k-mooney cdent: also i assume you are more or less the resouces that are available to work on it
16:04:42 cdent sean-k-mooney: I consider myself already fully booked
16:05:11 sean-k-mooney cdent: yep proably over booked
16:07:09 prometheanfire which db field is the 'volumes_attached' field sourced from?
16:07:36 prometheanfire smcginnis: you might know? ^
16:08:25 mriedem prometheanfire: block_device_mappings
16:08:42 smcginnis mriedem is the expert there. ;)
16:09:19 prometheanfire mriedem: thanks, I can't reproduce it, but cinder created a volume fine, but nova was 504 on the attach (it shows up in the db but no action taken on the compute host)
16:09:31 prometheanfire restart nova services with debug/verbose and it attaches fine
16:09:51 prometheanfire since it was only half added, the api can't delete it
16:10:28 mriedem prometheanfire: meaning nova-compute asked cinder to create the volume, right?
16:10:37 mriedem bdm source_type != 'volume'
16:11:21 mriedem "the api can't delete it" - you mean the cinder api can't delete the volume? or the compute api can't delete the server?
16:11:25 prometheanfire mriedem: no, volume created manually (worked), then nova was asked to attach volume (failed)
16:11:32 prometheanfire api can't remove it from the instance
16:11:41 prometheanfire the failed part half attached it to the instance
16:12:20 prometheanfire added to block_device_mapping table but no action taken on compute host
16:12:23 openstackgerrit Merged openstack/os-vif master: Add Python 3 Train unit tests https://review.opendev.org/669438
16:12:50 mriedem so there is a bdm, and the volume shows up for the server but when you try to detach it what happens? nova gets an error from cinder?
16:12:58 mriedem saying it's already detached (not attached) or something?
16:13:15 prometheanfire nova questions cinder if it's attached, cinder says no, nova does nothing
16:14:00 mriedem as in nova doesn't delete the bdm so it stops showing up as attached to the server
16:14:03 prometheanfire does nothing -> says it's not attached, why are you asking me to detach something that's not attached (returned to user)
16:14:15 mriedem which release?
16:14:19 prometheanfire stein
16:14:50 prometheanfire I haven't been able to reproduce this after restarting nova stuff infra side
16:15:01 prometheanfire so maybe some timeout or hanging connection?
16:15:20 mriedem my guess is cinder is failing on this call: https://github.com/openstack/nova/blob/86524773b8cd3a52c98409c7ca183b4e1873e2b8/nova/compute/api.py#L4175
16:15:43 prometheanfire maybe, for the detach part
16:16:14 mriedem well what error do you get from the compute API?
16:16:26 mriedem "Invalid volume: %(reason)s"
16:16:53 prometheanfire No volume with a name or ID of '3e92f9b6-1a3b-4a7e-8487-6ff253e888db' exists.
16:17:01 prometheanfire since I removed the volumes
16:17:36 prometheanfire Invalid volume: Invalid input received: Invalid volume: Unable to detach volume. Volume status must be 'in-use' and attach_status must be 'attached' to detach. (HTTP 400) (Request-ID: req-01d8a3c5-66f4-43a4-bac5-9f1104a292fe) (HTTP 400) (Request-ID: req-358132e4-38a1-493a-b87b-d135791bae2d)
16:17:40 mriedem msg = _("Unable to detach volume. Volume status must be 'in-use' "
16:17:41 mriedem yup
16:17:41 mriedem "and attach_status must be 'attached' to detach.")
16:17:48 mriedem that's the begin_detaching call failing
16:18:05 prometheanfire because it never actually attached :D
16:18:11 openstackgerrit sean mooney proposed openstack/os-vif master: Sync Sphinx requirement https://review.opendev.org/666387
16:18:14 mriedem right i get it
16:18:41 openstackgerrit Balazs Gibizer proposed openstack/nova master: nova-manage: heal port allocations https://review.opendev.org/637955
16:18:48 gibi mriedem: rebased ^^
16:19:43 mriedem prometheanfire: interesting you wouldn't hit this in the compute service when the attach fails and delete the bdm https://github.com/openstack/nova/blob/86524773b8cd3a52c98409c7ca183b4e1873e2b8/nova/compute/manager.py#L5722
16:20:41 prometheanfire haven't been able to reproduce after restart of services, but am leaving it in debug mode in case it happens again
16:21:06 openstackgerrit Balazs Gibizer proposed openstack/nova master: Move consts from neutronv2/api to constants module https://review.opendev.org/668945
16:21:07 mriedem were there errors in the compute log after the failed attach about not being able to delete the bdm?
16:21:14 openstackgerrit Balazs Gibizer proposed openstack/nova master: Use neutron contants in cmd/manage.py https://review.opendev.org/668946
16:21:26 openstackgerrit Balazs Gibizer proposed openstack/nova master: Add 'resource_request' to neutronv2/constants https://review.opendev.org/668947
16:22:04 prometheanfire compute didn't even attempt to attach from what I could see
16:22:12 prometheanfire didn't get that far

Earlier   Later