Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-05-17
12:42:53 mriedem artom: using mox to tweak existing tests that are using mox is acceptable
12:43:53 sfinucan Miouge: The comments in this spec should explain the position :) https://review.openstack.org/#/c/387295/
12:44:12 sfinucan ...and host aggregates would be preferred, yes
12:44:54 artom mriedem, fair enough. If I ever find the discussion I'm thinking of I'll wave it in your face, but yeah, I won't be lazy and do the new test with mock ;)
12:45:16 mriedem artom: i remember it, it was some nova-network live migration thing
12:45:29 mriedem and i believe dansmith went back afterward and re-wrote the tests in mock
12:46:08 mriedem artom: see #4 here https://docs.openstack.org/infra/manual/developers.html#peer-review
12:51:30 dane-fichter hey dansmith, could you take a look at my trusted certs patch when you get a chance?
12:51:37 dane-fichter https://review.openstack.org/#/c/457711/
12:52:26 dane-fichter It's passing unit tests, not sure if there's anything else I need to implement here
12:53:36 sdague dansmith / mriedem: thinking about the thing that came up in the cells forum session around searching by ip ....
12:53:51 sdague it's really pathologically bad in our code
12:54:18 mriedem i looked it up after the session,
12:54:23 mriedem and it wasn't mysql or pg specific
12:54:30 sdague nope
12:54:31 mriedem it was originally just about a security bug
12:54:39 sdague mriedem: what was the security bug?
12:55:00 mriedem api would call down to nova-network which would pull all instances from the db and iterate them filtering on ip, something like that
12:56:39 sdague ok, so removing nova-network out of the flow, we're now just loading every instance, loading every network info cache, and grepping through each json?
12:56:46 sdague especially in the all_tenants case
12:57:49 mriedem https://github.com/openstack/nova/commit/ae781ee97947c33d6d43e4c21df4f338c875bf1c was the original change
12:58:43 mriedem sdague: not sure how you'd do the filtering in the db when the instance_info_cache.network_data is serialized json
12:59:14 mriedem *network_info
12:59:30 sdague right, in order to make it efficient we'd need to extract the search keys that we wanted to hit into a column
12:59:48 sdague which you could do, just denormalize it a bit
13:00:55 sdague given that the ip search is regex, you could even just make it a comma separated list, and still do the second stage filter, but it would drop a lot of db bloat
13:01:37 sdague anyway, just pondering that, given it seemed like it was a query many wanted to do, and it's one of our most pathologically bad ones
13:01:51 mriedem and it's also been this way forever..
13:02:26 mriedem if we had numbers on how long it takes to do the filter on an all_tenants call with 1000 instances or something, then it might be more interesting
13:02:53 sdague mriedem: yeh data would be good
13:03:22 sdague it does also mean though if you exceed 1000 instances in your filter set before ips, you'll get the wrong answer, right?
13:03:32 sdague that's actually the thing I'm more concerned about
13:03:54 mriedem true, if the ip you're looking for is the 1001 instance, you're not going to get anything back
13:04:15 sdague if nothing else, we should describe that issue
13:05:15 mriedem yeah, definitely,
13:05:19 mriedem could go into the api-ref
13:05:44 mriedem i feel like there was something else we said we'd doc in the api-ref but i'm forgetting - i think notes are in etherpads though, and i haven't digested those yet
13:06:25 mriedem ip (Optional) query string An IPv4 address to filter results by.
13:06:25 sdague well, in mysql we could - https://dev.mysql.com/doc/refman/5.7/en/json-search-functions.html#function_json-contains
13:06:32 mriedem ^ is the api-ref and doesn't mention anything about this
13:06:40 sdague mriedem: yep, I know
13:06:46 sdague I honestly didn't realize how bad it was
13:07:59 mriedem looks like postgresql does too https://www.postgresql.org/docs/9.3/static/functions-json.html
13:08:27 mriedem http://schinckel.net/2014/05/25/querying-json-in-postgres/
13:09:52 rabel Hi there, I'm a little bit confused about quotas in nova. To start with: Can anyone tell me, what a quota class is? And how are those related to default quotas? How do default quotas work in detail? For example why does a project have default quotas? I thought default quotas are global quotas that take effect for projects without quotas?
13:12:06 mriedem rabel: se
13:12:07 mriedem *sec
13:12:37 mriedem rabel: some notes here https://docs.openstack.org/developer/nova/quotas.html
13:12:52 mriedem rabel: there is only one quota class and it's "default"
13:13:18 mriedem some more info here https://review.openstack.org/#/c/411035/
13:14:47 rabel mriedem: thank you very much! do you know if other openstack components differ in this regard?
13:18:35 mkoderer mriedem: ping
13:18:40 bhagyashris Hi all, Can anyone please guide me How to check the memory consumption or memory during the process initialization suppose for n-cpu service? Is there any command to check this?
13:22:10 rabel mriedem: what is still confusing to me is the --default option of "openstack quota show", which shows default quotas for a given project or class. that does not seem to make a lot of sense. what do you think?
13:30:35 mriedem rabel: i think that's trying to match the nova quota-defaults CLI
13:31:38 mriedem mkoderer: pong
13:32:10 mkoderer mriedem: dgonzalez and I wanted to discuss a tricky bug with you: https://bugs.launchpad.net/nova/+bug/1691131
13:32:11 openstack Launchpad bug 1691131 in OpenStack Compute (nova) "IO stuck causes nova compute agent outage" [Undecided,New]
13:32:55 mkoderer nova-compute stops sending it's heartbeats if we have a NFS overload
13:33:30 mkoderer and with this we can't control the cloud anylonger since stopping of vms is not possible anylonger
13:33:41 mriedem mkoderer: which release?
13:33:56 mkoderer we are on liberty - but the code is the same on master
13:34:05 mriedem "The nova-compute services scans over all attached volume paths"
13:34:11 mriedem which part specifically are you referring to there?
13:34:40 rabel mriedem: so, let me summarize: the only quota class is "default" and the concept of quota classes will probably soon be removed. default quotas are quotas that take effect if there is no project-specific quotas. and having default quotas for a project/class does not make sense. is this correct in your opinion?
13:35:24 mriedem "and having default quotas for a project/class does not make sense" - why does that not make sense?
13:35:34 mriedem rabel: having default quota (limits) is fine
13:36:08 mriedem having an api and database for quota classes when there is only ever a single entry though, is weird
13:36:15 mriedem *database table
13:36:57 mkoderer mriedem: this part blocks: https://github.com/openstack/nova/blob/902b7bf6f5425824a0661b8e4beac4a894749c03/nova/virt/libvirt/driver.py#L7022
13:37:28 rabel mriedem: default quotas do make sense. but how are they connected to a specific project? isn't the idea of default quotas that they are global?
13:37:28 dansmith sdague: mriedem: setting the precedent of loading json in mysql seems not awesome
13:37:42 dansmith sdague: mriedem: I'd almost rather just build a separate index of that thing, if it's a thing people do often
13:38:06 dansmith sdague: mriedem: also, it's something rather easy that a client can do right? find the port/device_id in neutron and then show the instance in nova?
13:41:57 bauzas dansmith: edleafe: tbc, I'm working on updating my series given dan's comments
13:42:14 dansmith bauzas: sweet :)
13:42:33 bauzas just has a problem with a 2-instances request
13:42:42 edleafe bauzas: cool. I'll be pushing an update to my series, but it's just a pep8 fix, so don't freak out
13:42:42 bauzas because I need more hosts than just 4
13:42:56 sdague dansmith: expect, neutron api isn't always exposed correctly directly to users
13:43:17 sdague dansmith: yeh, secondary index would be fine as well
13:43:36 bauzas edleafe: which one do you think you want to update ? the ones about passing the UUIDs and the scheduler claims?
13:44:26 edleafe bauzas: yep
13:44:30 bauzas ok
13:45:19 dansmith sdague: could we also just special case that search by asking neutron ourselves and then pulling the relevant instances out ?
13:45:33 openstackgerrit OpenStack Proposal Bot proposed openstack/nova master: Updated from global requirements https://review.openstack.org/465283
13:45:35 sdague dansmith: they support search by regex?
13:46:02 sdague I think it gets complicated when there are multiple filters
13:46:15 sdague but especially what we allow in our filters today
13:46:24 dansmith sdague: I dunno, they'd have to for the client approach to work, but I didn't realize that searching by regex was a thing, I thought the use-case was "find me the instance with this ip"
13:47:06 sdague it's definitely a regex today, which is kind of suck
13:47:09 sdague but that's what it is
13:47:14 sdague and has been
13:48:43 mriedem mkoderer: that's in a ploop block, are you using ploop?
13:49:14 mriedem mkoderer: is there a way to detect a bad mountpoint with nfs?
13:53:10 dgonzalez mriedem: mkoderer sent the wrong line, the correct line is https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L7025 (below the loop block)
13:53:31 openstackgerrit Béla Vancsics proposed openstack/nova master: Transform keypair.delete notification https://review.openstack.org/463946
13:53:50 dgonzalez mriedem: But both lines basically have the same content, so I guess the issue could occur in both blocks
13:54:53 mriedem i'm going to be away for a couple of hours
13:58:49 alex_xu johnthetubaguy: sdague it would be great to merge this one first when you have free time, it can reduce some conflict in later patch. thanks in advance

Earlier   Later