Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-09-25
17:46:32 melwitt sean-k-mooney: I don't know much about that but the expected type will be different depending on whether the cell returned a result or an exception
17:47:04 tssurya melwitt: for me the requirement came because of wanting to differentiate between InstanceNotFound valid exception versus an actuall cell down exception here: https://review.openstack.org/#/c/591658/7/nova/compute/api.py@2327 , feel free to leave a comment and we can also raise it at tomorrow's cells meeting; need to leave office now
17:47:58 sean-k-mooney in c++ std::expect models a type that either pacages the expected result or and error type. when you use it you can test if it it has an error or not with an if expect ...
17:48:22 melwitt tssurya: ok, will look there. I was thinking we would need to be able to differentiate exceptions but didn't see a concrete example till now
17:48:50 melwitt sean-k-mooney: oh, nice
17:49:37 AJaeger nova stable team, could you review a simple zuul job change, please? https://review.openstack.org/602018 https://review.openstack.org/602019 https://review.openstack.org/602022 https://review.openstack.org/602023
17:49:45 AJaeger mriedem: I answered your comment on https://review.openstack.org/#/c/602018/1/.zuul.yaml
17:49:54 sean-k-mooney melwitt: so my preference would be have result be a type that you can just do "if result: #handel sucess else : # handel error"
17:50:58 melwitt sean-k-mooney: we also have "did not respond" i.e. timeout waiting for result
17:51:02 sean-k-mooney melwitt: to do that you would have to override __bool__ in the wrapper class to see if the value was an exception type or not
17:51:19 melwitt which could be an error, but then you have to check the type of the error
17:51:34 melwitt but I guess we have to do that anyway if we're differentiating
17:52:46 sean-k-mooney melwitt: ya. if you stuff the exeption into the value field you can just rerais it and cactch it polymorifclly assuming the timeout is a different exception type
17:53:13 melwitt yeah, could do that
17:53:49 sean-k-mooney melwitt: so basically its the same as what you suggted but your hiding the type checking in a dedicated class that does it for you instead of doing it in multiple places
17:55:34 sean-k-mooney melwitt: python3's concurrent futures basically wrap this up nicly for you too but they are python 3 only https://docs.python.org/3/library/concurrent.futures.html#concurrent.futures.Future
17:56:09 melwitt sean-k-mooney: thanks
17:59:57 sean-k-mooney melwitt: want me to mock up an example of using what i described so you can see how it would work. im not sure its the correct solution in this case but its what i would try personally.
18:05:09 melwitt sean-k-mooney: an example that would work for python2 also? I would be interested in looking at it if you mock it up
18:05:57 sean-k-mooney melwitt: yes i can make it work for python too. ill mock something up and push it with some unit tests showing its usage later tonight
18:06:17 sean-k-mooney *python 2
18:06:50 melwitt thanks
18:16:34 openstackgerrit Jack Ding proposed openstack/nova master: Handle missing marker during online data migration https://review.openstack.org/605164
18:17:11 jaypipes man I fucking hate the scheduler unit tests.
18:17:15 jaypipes what a pile of shit.
18:17:59 sean-k-mooney jaypipes: let me guess asserting behavior 3 levles deep then just the local function behvior?
18:19:39 sean-k-mooney anyone know where the devstack log is stored by default? ther eused to be an xstack log in /opt/stack/logs but that seams to not be a thing anymore
18:20:10 jaypipes sean-k-mooney: more the endless hierarchy of copy/pasted test case classes.
18:20:39 sean-k-mooney jaypipes: yeah OOP
18:22:11 sean-k-mooney jaypipes: did i mention im playing with creating a programming language. it "object orianted" current it does not allow inheritence but you can have objects :P
18:26:30 AJaeger mriedem: could you review the other stable backports of that ceph change as well, please?
18:26:47 mriedem yeah
18:28:22 AJaeger thanks
18:31:58 melwitt what is going on with the gate lately? changes that have been in the queue for 26 hours
18:33:12 AJaeger melwitt: high failure rate plus one of our clouds down, see clark's email from a week ago or so
18:33:49 melwitt found it, thank you
18:33:52 AJaeger melwitt: http://lists.openstack.org/pipermail/openstack-dev/2018-September/134867.html
18:34:40 AJaeger melwitt: still applies regarding cloud down - clarkb is testing a fix right now
18:46:27 openstackgerrit Jonte Watford proposed openstack/nova master: This commit adds __str__ and __repr__ methods to objects InstanceNUMACell and NUMACell. These will generate more readable representations of the objects for use in logs. https://review.openstack.org/600269
18:48:57 mriedem AJaeger: waiting to see the renamed job run on ocata https://review.openstack.org/#/c/602023/
18:49:44 AJaeger mriedem: good catch - you won't see it. We miss the parent devstack job there. This needs different handling - I'll fix...
18:54:27 openstackgerrit Jonte Watford proposed openstack/nova master: This commit adds __str__ method to objects InstanceNUMACell and NUMACell. These will generate more readable representations of the objects for use in logs. https://review.openstack.org/600269
19:01:36 mriedem jaypipes: i've got an o.vo problem that maybe you can help solve,
19:01:58 mriedem i've got an Instance object from cell1, pulled fresh from the db so there are no fields marked as 'changed',
19:02:30 mriedem i want to clone that object and re-create it in cell2 which should mean just changing the context to point at cell2 db and delete the id field so Instance.create() won't puke on it,
19:03:01 mriedem the problem is, the cloned object doesn't have any changed fields marked, so Instance.create() doesn't actually save anything into the cell2 db for the instance for things like vm_state, task_state, etc
19:03:54 mriedem know of any way to dirty up the object so it looks new? I tried a obj_from_primitive(obj.obj_to_primitive()) dance but that doesn't work b/c it maintains the 'changes' list of dirty fields
19:04:30 mriedem i'm thinking i might just have to do something like, new_obj = objects.Instance(new_context, **instance.obj_to_primitive()['data'])
19:05:58 melwitt mriedem: I think the only precedent we have for that is the get_new_instance in build request https://github.com/openstack/nova/blob/master/nova/objects/build_request.py#L237
19:07:29 sean-k-mooney mriedem: can you do a deep copy and then instead update the cell and then instaead of resetting the fileds just mark them as dirty?
19:07:30 mriedem blech, ok i'll hack that up the same
19:08:07 melwitt there might be a better way to do it, just mentioning that as a data point
19:08:49 mriedem yeah i'll copy that,
19:08:55 mriedem as i've seen and from dansmith's comment in that code,
19:09:01 mriedem keeping the fields dirty would involve ovo internals
19:09:24 melwitt aye
19:09:32 mriedem like passing a dirty_white_boy=True kwarg to obj_from_primitive
19:19:14 jaypipes mriedem: yeah, it's blech.
19:20:38 mriedem i'm still waiting for someone to get my foreigner reference
19:24:24 jaypipes heh
19:24:29 melwitt sorry, I only know the major foreigner hits
19:26:02 mriedem hey it made it to the top 12 https://en.wikipedia.org/wiki/Dirty_White_Boy_(song)#Chart_performance
19:26:36 melwitt huh.
19:37:28 sean-k-mooney yeah cirros does not have sriov drivers for 1gb intel nics ...
19:38:07 mriedem ok i've got something hacked up for now
19:38:28 mriedem http://paste.openstack.org/show/730743/
19:39:41 sean-k-mooney mriedem: without messing with internal that looks resonably clean
19:40:17 sean-k-mooney mriedem: it would be nice to a "mark_dirty" function instead of setattr(new_obj, field, getattr(obj, field)) but that works
19:40:53 mriedem i'll post something to ovo that dan can look at while he's pooing in a cave
19:50:18 openstackgerrit Matt Riedemann proposed openstack/nova master: Option "scheduler_default_filters" is deprecated. https://review.openstack.org/604148
19:51:35 imacdonn mriedem: is there an easy way to see what 'nova-manage db online_data_migrations' would do, without actually doing it ?
19:51:37 sean-k-mooney melwitt: im going to call it a night an grab a beer. am im going to start hacking on that exection wrapper class and ill push something up in an hour or so while i wait for dinner. ill add you as a reviewer when i submit the patch.
19:52:46 mriedem imacdonn: as in a dry run option?
19:52:49 mriedem no there isn't
19:53:02 mriedem jaypipes: sean-k-mooney: like this https://review.openstack.org/605199
19:53:19 melwitt sean-k-mooney: cool, thanks
19:53:32 imacdonn mriedem: yeah. ok. I think that would be useful. I tried hacking out the check for max-count being greater than zero, but it's not that simple, apparently
19:54:27 mriedem imacdonn: well, you'd have to plumb that down through all of the online data migration routines
19:54:31 mriedem to not actually commit any changes
19:54:34 mriedem so it's not trivial
19:54:49 imacdonn mriedem: also, this doesn't seem right. The counts in the table should be non-zero? Caveat: it's Queens, so if anything's been fixed since then .... http://paste.openstack.org/show/730819/
19:55:30 sean-k-mooney mriedem: is the condition on line 401 reversed? e.g. should it be if leave_dirty?
19:57:34 sean-k-mooney mriedem: i would have expected the changed filed to be the dirty fields personally but i have not really looked at how this works closely
19:58:15 mriedem no
19:58:46 mriedem if not False (True): set the same changed fields on the copy as were on the primitive
19:58:53 mriedem which defaults to no changed fields
19:59:02 openstackgerrit Jonte Watford proposed openstack/nova master: Add __str__ method to objects NUMACell instances https://review.openstack.org/600269
20:01:07 sean-k-mooney mriedem: i have added myself as a review and ill step through it properlly tomorow.
20:01:07 openstackgerrit Merged openstack/nova stable/rocky: Follow devstack-plugin-ceph job rename https://review.openstack.org/602018
20:01:27 mriedem imacdonn: well that table definitely looks wrong...
20:01:46 imacdonn mriedem: yeah. Trying to figure out how that could happen
20:02:05 mriedem my guess is you're dumping the last result set,
20:02:11 mriedem which is when there were no more things to migrate
20:02:18 mriedem so you get 0s at the end
20:02:48 imacdonn I'm just running the command :)
20:02:54 imacdonn but yeah
20:02:59 mriedem you said you hacked something
20:03:19 imacdonn Oh, I only hacked out the check that max-count has to be >0 .. but I'm not applying that here
20:03:22 mriedem however,
20:03:23 mriedem http://logs.openstack.org/87/604687/4/check/neutron-grenade/6c10d43/logs/grenade.sh.txt.gz

Earlier   Later