Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-03-17
16:17:46 cfriesen and other stuff like cburgess pointed out
16:17:56 cburgess mriedem cfriesen But a change to have conductor/compute cast to/from each other and simply take action once the call comes in rather then waiting on it and forcing it to take some fixed amount of time is a way way better solution.
16:18:23 cburgess cfriesen Right... which is why I think moving everything to cast calls to/from conductor is way better.
16:18:31 cfriesen cburgess: mriedem: agreed...would we need to change the handling of the overall migration completion timeout? (checking code)
16:18:56 cburgess cfriesen Not sure if that timer starts during pre flight.
16:19:07 mriedem the timer is while it's moving data right?
16:19:14 cfriesen cburgess: mriedem: looks like the migration completion timeout is just a libvirt thing
16:19:15 mriedem to kill it if it gets stuck
16:19:17 cburgess mriedem I thought that timer was going away?
16:19:23 mriedem cburgess: i believe it is
16:19:27 cfriesen cburgess: the progress timer is going away
16:19:31 mriedem johnthetubaguy and the osic fok are looking at that
16:19:32 cburgess Ahhh
16:19:33 mriedem *folk
16:19:58 johnthetubaguy sorry, I missed all that chat there
16:20:11 cburgess johnthetubaguy Welcome to the party!
16:20:38 cburgess johnthetubaguy Wait its after 4PM on a Friday there... why are you at the pub?
16:20:47 cfriesen my concern about just moving to cast is that if one of those gets lost we're stuck with a migration in progress
16:20:47 cburgess Sorry why aren't you
16:20:59 cburgess cfriesen Yes thats the rub.
16:21:02 johnthetubaguy heh, I am around for a few hours yet
16:21:17 johnthetubaguy got band rehersal before I get close to a pub anyways!
16:21:20 cburgess cfriesen We would have to be able to cancel it. Could the migration records accomplish that?
16:21:24 cdent figleaf I think that's inevitable unless we want solely use the external interface of the compute manager (which might be fine), but even with there would still need to be something to causing the period job to run, and to call the external interface
16:21:29 mriedem cfriesen: cburgess: so i just looked at oslo.messaging, and yeah it looks trivial to pass a new timeout in when constructing the client
16:21:42 cfriesen I wonder if the existing live migration abort API would be sufficient
16:21:59 cburgess mriedem I don't even think you have to do it when constructing the client. I think with an existing client its just an arg you can pass on call.
16:21:59 figleaf cdent: yeah, I guess I'm just too idealistic :)
16:22:05 johnthetubaguy mriedem: I thought that was a per call thing we can tweak?
16:22:14 mriedem cburgess: oh that's even better b/c we cache the clients
16:22:21 johnthetubaguy yeah, what cburgess said
16:22:22 cdent figleaf: no, you just are trying to apply principles that might work elsewhere, here
16:22:34 cburgess cfriesen Right thats what I'm thinking. If we know the state is "we made the cast to run pre live migration" then we should be able to cancel that by making the right call to abort it.
16:22:36 johnthetubaguy cburgess: I am +1 increasing the timeout for the lots of volumes case
16:22:44 figleaf cdent: like I said, too idealistic
16:23:04 superdan mriedem: no caching of the client anymore right?
16:23:09 superdan well, for compute anyway
16:23:23 cburgess johnthetubaguy Some type of equation you thjink? Before call is made look at size of BDM and make some type of auto-adjust? Or just a new config value for pre-live timeout?
16:23:31 cfriesen mriedem: I think there's an example of the per-call timeout in nova/cells/rpcapi.py
16:23:34 superdan well, sometimes in compute actually
16:23:38 johnthetubaguy cburgess: just a large one I think
16:23:51 mriedem yeah i see the timeout kwarg to the client.prepare method
16:23:53 johnthetubaguy cburgess: its only there as a "oh dear, you must have given up" kind of timeout
16:23:58 johnthetubaguy mriedem: yeah, thats the one
16:24:04 cburgess johnthetubaguy Just an arbitrary "This call can suck, wait 10 minutes" type thing?
16:24:11 johnthetubaguy mriedem: I thought we used that already for something, I can't remember now
16:24:13 superdan it'd be nice if we had a per-call heartbeat mechanism,
16:24:17 superdan which o.msg could provide for us
16:24:27 superdan send a "still working" reply to the reply queue which would reset the timer
16:24:29 cburgess johnthetubaguy The timeout would be a stop gap really. I think mriedem and cfriesen idea about moving to cast is the right solution but probably going to take more effort.
16:24:31 johnthetubaguy superdan: thats the "task" framework in my head
16:24:37 cburgess superdan OMG yes
16:24:40 superdan johnthetubaguy: no, I mean something at the o.msg layer
16:24:47 superdan johnthetubaguy: we could solve it in nova with a task too
16:24:50 johnthetubaguy superdan: yeah, check if job is progressing, and you know the owner
16:24:54 mriedem def ping(self, context, arg, timeout=None):
16:24:54 mriedem arg_p = jsonutils.to_primitive(arg)
16:24:54 mriedem cctxt = self.client.prepare(timeout=timeout)
16:24:57 mriedem we do have something using it
16:25:00 superdan but it'd be nice for lots of things if o.msg would heartbeat a particular call
16:25:11 johnthetubaguy superdan: oh I see what you mean now
16:25:29 cfriesen mriedem: you can set it in the client init but also in the cctxt.call()
16:25:32 superdan johnthetubaguy: you could just send something to the reply queue like "this isn't the reply, but I ain't dead yet"
16:25:50 cburgess mriedem cfriesen johnthetubaguy superdan: So 1 bug and 1 spec here? bug = set a "large" timeout for the pre_live call today, spec = move to casting?
16:25:50 mriedem cfriesen: i don't actually see the timeout kwarg in the rpc client call() method in oslo.messaging
16:25:51 johnthetubaguy superdan: yeah, like a "continue" or whatever
16:25:53 mriedem it's in the prepare() method
16:26:18 superdan johnthetubaguy: or lease renewal
16:26:31 mriedem cfriesen: http://git.openstack.org/cgit/openstack/oslo.messaging/tree/oslo_messaging/rpc/client.py#n178
16:26:45 mriedem cfriesen: call() doesn't use a timeout kwarg http://git.openstack.org/cgit/openstack/oslo.messaging/tree/oslo_messaging/rpc/client.py#n151
16:27:16 cfriesen mriedem: we're using it in nova/cells/rpcapi.py, line 332
16:27:40 johnthetubaguy superdan: yeah, reset the timeout timer
16:27:45 mriedem cfriesen: that looks like a mistake
16:28:06 cburgess mriedem cfriesen Yeah the kwargs just gets pased all the way down and I think that one gets dropped.
16:28:08 mriedem cfriesen: the oslo.messaging rpc client call() method takes any **kwargs
16:28:16 mriedem right, the call() method doesn't use the timeout
16:28:18 mriedem the prepare() method does
16:28:32 mriedem which is nova/cells/rpcapi.py L326
16:28:48 cburgess mriedem Right.
16:29:08 cburgess mriedem I still suspect its doable... we just have to modify things a bit more to make it work.
16:29:18 cfriesen mriedem: I think maybe the call() routine used to take a timeout
16:29:23 mriedem maybe
16:29:27 mriedem anyway yeah this is easy
16:29:48 mriedem https://github.com/openstack/nova/blob/master/nova/compute/rpcapi.py#L720
16:29:59 mriedem you use your nova config pre_lie_migration_timeout value there
16:29:59 cburgess https://github.com/openstack/nova/blob/master/nova/compute/rpcapi.py#L720
16:30:01 cburgess LOL yeah
16:30:08 cburgess mriedem Just found that same line.
16:30:09 mriedem default to None to get the default from oslo.messaging
16:30:14 mriedem bump to 10 minutes if you need to
16:30:18 cburgess mriedem Yup
16:30:23 cburgess mriedem So fixed value or config option?
16:30:27 mriedem config
16:30:33 cburgess Where is sdague to yell at me about config options?
16:30:37 cburgess mriedem Bug? or Spec?
16:30:41 mriedem this is completely operator side config
16:30:48 mriedem this is a bug i think
16:31:01 mriedem the spec is changing the blocking call to two casts
16:31:07 cburgess mriedem Right
16:31:27 cburgess mriedem OK... I shall file bug with code since this is easy and allow me to get ATC again. :P

Earlier   Later