| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul-jobs] 866945: Avoid tox-bindep in test-tox job https://review.opendev.org/c/zuul/zuul-jobs/+/866945 | 00:00 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866947: ensure-kubernetes: drop kubic repos https://review.opendev.org/c/zuul/zuul-jobs/+/866947 | 00:15 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866947: ensure-kubernetes: drop kubic repos on 22.04 https://review.opendev.org/c/zuul/zuul-jobs/+/866947 | 00:15 | |
| -@gerrit:opendev.org- Clark Boylan proposed: [zuul/zuul-jobs] 866943: Flip ensure-tox to tox v4 by default https://review.opendev.org/c/zuul/zuul-jobs/+/866943 | 00:25 | |
| -@gerrit:opendev.org- Clark Boylan proposed on behalf of James E. Blair https://matrix.to/#/@jim:acmegating.com: [zuul/zuul-jobs] 866945: Avoid tox-bindep in test-tox job https://review.opendev.org/c/zuul/zuul-jobs/+/866945 | 00:25 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/nodepool] 866948: nodepool-functional-k8s: use docker backend https://review.opendev.org/c/zuul/nodepool/+/866948 | 00:47 | |
| @jim:acmegating.com | ianw: Clark perhaps we should consider using kind instead of minikube. that's how i do all my local testing. | 00:49 |
|---|---|---|
| @iwienand:matrix.org | probably if we want the thing that works the best on ubuntu it would be microk8s | 00:51 |
| @iwienand:matrix.org | it looks like kind uses containerd/docker by default? i guess that's pretty close to minikube with a docker-backend | 00:53 |
| -@gerrit:opendev.org- Clark Boylan proposed: [zuul/zuul-jobs] 866943: Flip ensure-tox to tox v4 by default https://review.opendev.org/c/zuul/zuul-jobs/+/866943 | 00:53 | |
| -@gerrit:opendev.org- Clark Boylan proposed on behalf of James E. Blair https://matrix.to/#/@jim:acmegating.com: [zuul/zuul-jobs] 866945: Avoid tox-bindep in test-tox job https://review.opendev.org/c/zuul/zuul-jobs/+/866945 | 00:53 | |
| @iwienand:matrix.org | in so far as both pull docker from dockers repos | 00:53 |
| @jim:acmegating.com | ianw: yeah, it's k8s inside of docker containers, so the only host-level requirement is docker (i haven't tried it with podman) | 00:54 |
| @clarkb:matrix.org | I have no objections to kind. I think the main reason to minikube was its ubiquity but maybe that is no longer the case as the tooling under it fragments | 00:57 |
| @jim:acmegating.com | ianw: i like the seeming system-independence and ease of install of kind (i feel like it ought to work equally well despite base os), and also it seems pretty close to a vanilla k8s (you can even run it multi-node -- on a single node), but i'm not stuck on it. :) microk8s says it's "conformant" so maybe it's okay. :) | 00:57 |
| @michael_kelly_anet:matrix.org | I use microk8s on my pi cluster - as far as I can tell it's just a regular k8s with some optional builtins. | 01:25 |
| @michael_kelly_anet:matrix.org | And it's a `snap` to install ;) | 01:25 |
| @michael_kelly_anet:matrix.org | The ease of use (on Ubuntu) is the main reason I use it in lieu of something like k3s | 01:26 |
| @iwienand:matrix.org | it seems like ensure-kubernetes is installing cri-o even when the docker-backend is selected. i think that's my bug | 01:27 |
| @iwienand:matrix.org | > <@michael_kelly_anet:matrix.org> And it's a `snap` to install ;) | 01:29 |
| there's no point tilting at windmills about snaps, and probably for this case where you never want to touch anything it's fine, but something about them as the base form of software distribution is annoying | ||
| @michael_kelly_anet:matrix.org | hah | 01:30 |
| @michael_kelly_anet:matrix.org | I generally find `snap` to be an awful distribution system, fwiw. Especially for desktop apps. | 01:30 |
| @michael_kelly_anet:matrix.org | Otoh, it was nice in this case because it makes provisioning and managing nodes a bit easier. | 01:30 |
| @michael_kelly_anet:matrix.org | Probably would have been better to be a proper .deb | 01:31 |
| @michael_kelly_anet:matrix.org | the "`snap` to install part" was supposed to be a bit of wordplay though :) | 01:32 |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866950: ensure-kubernetes: don't install cri-o with docker backend https://review.opendev.org/c/zuul/zuul-jobs/+/866950 | 01:39 | |
| -@gerrit:opendev.org- Clark Boylan proposed: [zuul/zuul-jobs] 866943: Flip ensure-tox to tox v4 by default https://review.opendev.org/c/zuul/zuul-jobs/+/866943 | 01:49 | |
| -@gerrit:opendev.org- Clark Boylan proposed on behalf of James E. Blair https://matrix.to/#/@jim:acmegating.com: [zuul/zuul-jobs] 866945: Avoid tox-bindep in test-tox job https://review.opendev.org/c/zuul/zuul-jobs/+/866945 | 01:49 | |
| @iwienand:matrix.org | sigh, using the docker backend requires critool --- which comes from --- the suse kubic repos. we can download a binary directly, i'll try that | 01:54 |
| @clarkb:matrix.org | one upside to using kind is we can use the same tool on multiple platforms | 01:54 |
| @clarkb:matrix.org | which for zuul library things is ideal | 01:54 |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866950: ensure-kubernetes: don't install cri-o with docker backend https://review.opendev.org/c/zuul/zuul-jobs/+/866950 | 01:55 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866950: ensure-kubernetes: don't install cri-o with docker backend https://review.opendev.org/c/zuul/zuul-jobs/+/866950 | 02:05 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866950: ensure-kubernetes: don't install cri-o with docker backend https://review.opendev.org/c/zuul/zuul-jobs/+/866950 | 02:25 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866950: ensure-kubernetes: don't install cri-o with docker backend https://review.opendev.org/c/zuul/zuul-jobs/+/866950 | 02:51 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 03:54 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 03:56 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 03:58 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 04:08 | |
| @iwienand:matrix.org | i can't seem to get minikube to work without the suse CRI packages. i feel like it *might* be able too, but i'm not sure how many hours it is worth in effort. i'm sure suse will fix their key soon enough; but also, they've deprecated distribution (https://kubic.opensuse.org/blog/2022-06-10-kubic-retired/) this doesn't seem a solid foundation. since microk8s is the "official" ubuntu way, that seems like stop 1. we can investigate others too, but it would be nice to have something working | 04:10 |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 04:13 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 04:19 | |
| -@gerrit:opendev.org- Dong Zhang proposed: [zuul/nodepool] 866520: Block device mappings should apply for that label https://review.opendev.org/c/zuul/nodepool/+/866520 | 04:53 | |
| -@gerrit:opendev.org- Dong Zhang proposed: [zuul/nodepool] 866520: Block device mappings should apply for that label only https://review.opendev.org/c/zuul/nodepool/+/866520 | 04:57 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 05:07 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 05:20 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 05:33 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/nodepool] 866955: [wip] k8s microk8s https://review.opendev.org/c/zuul/nodepool/+/866955 | 05:36 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 06:02 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 07:36 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/nodepool] 866955: [wip] k8s microk8s https://review.opendev.org/c/zuul/nodepool/+/866955 | 07:38 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: [wip] ensure-kubernetes-microk8s https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 07:48 | |
| -@gerrit:opendev.org- Felix Edel proposed: [zuul/zuul] 866863: Store paused events on the build and report them https://review.opendev.org/c/zuul/zuul/+/866863 | 09:33 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: ensure-kubernetes: add microk8s support https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 09:38 | |
| -@gerrit:opendev.org- Felix Edel proposed: [zuul/zuul] 866863: Store paused events on the build and report them https://review.opendev.org/c/zuul/zuul/+/866863 | 09:39 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: ensure-kubernetes: add microk8s support https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 09:44 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/nodepool] 866955: nodepool-functional-k8s: switch to microk8s https://review.opendev.org/c/zuul/nodepool/+/866955 | 09:44 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/nodepool] 866955: nodepool-functional-k8s: switch to microk8s https://review.opendev.org/c/zuul/nodepool/+/866955 | 09:50 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: ensure-kubernetes: add microk8s support https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 09:52 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/nodepool] 866955: nodepool-functional-k8s: switch to microk8s https://review.opendev.org/c/zuul/nodepool/+/866955 | 10:11 | |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/zuul-jobs] 866953: ensure-kubernetes: add microk8s support https://review.opendev.org/c/zuul/zuul-jobs/+/866953 | 10:13 | |
| @iwienand:matrix.org | Clark: corvus https://review.opendev.org/q/topic:ensure-k8s-microk8s implements microk8s. as i pushed it the last time, all the jobs started working -- so kubic must have fixed their key up. but having spent more than a bit of time struggling with minikube, i think this has value. it's also revealed that k8s 1.24 doesn't work due ot something to do with service accounts and tokens, something i can look into | 10:27 |
| -@gerrit:opendev.org- Fabien Boucher proposed: [zuul/zuul] 851008: Improve check for event change_url attribute https://review.opendev.org/c/zuul/zuul/+/851008 | 10:48 | |
| -@gerrit:opendev.org- Fabien Boucher proposed: [zuul/zuul] 862678: Ensure a buildset_doc have a start_date https://review.opendev.org/c/zuul/zuul/+/862678 | 10:51 | |
| @clarkb:matrix.org | Looks like my last patchset to the zuul-jobs tox v4 stack last night did end up addressing all of the existing testing issues. I'll go ahead and WIP the one that changes the default to v4 to prevent early merging but those should all be ready for review and appropriate scheduling/planning/merging | 15:48 |
| @jim:acmegating.com | Clark: that starts at https://review.opendev.org/866939 and has +2s from me | 15:49 |
| @jim:acmegating.com | zuul-maint: ^ | 15:50 |
| @clarkb:matrix.org | after breakfast/morning things my goal is to look into the nodepool unittesting failures today. Then look into openstacksdk 0.103 statsd lack of info in latest nodepool launchers. | 15:51 |
| @jim:acmegating.com | i updated https://etherpad.opendev.org/p/6wky6mkVu4uo2c3vaLpJ to increase the date by one day and can send that out after we merge the change | 15:51 |
| @clarkb:matrix.org | sounds good | 15:52 |
| @clarkb:matrix.org | corvus: the issue with nodepool testing is the reregistering extra nodes problem again. We send a static node through a ready -> used -> delete cycle to force it to be reregistered and when that occurs we're registering it twice so that two node entries are valid allowing two instead of one node request to be fulfilled. https://a378786df9574e363fce-f29e759a440a8c469e5909803b48c54b.ssl.cf5.rackcdn.com/862630/6/gate/nodepool-tox-py38/303ef23/testr_results.html | 16:32 |
| @clarkb:matrix.org | its a race between syncing the main config and reregistering a deleted node | 16:32 |
| @jim:acmegating.com | Clark: why are we syncing the main config? | 16:33 |
| @clarkb:matrix.org | corvus: thats what the static driver does iirc to know when to register new server or unregister old ones | 16:33 |
| @jim:acmegating.com | right i mean what's prompting it | 16:33 |
| @jim:acmegating.com | what is the other actor in the race (one actor is the delete callback, right?) | 16:34 |
| @clarkb:matrix.org | its the cleanupLeakedResources periodic call I think | 16:34 |
| @jim:acmegating.com | gotcha | 16:35 |
| @clarkb:matrix.org | I think my idling check only handled a subset of this issue when we load a new config, but didn't handle the case where we just periodically cleanup resources? | 16:35 |
| @jim:acmegating.com | Clark: there is a mutex around all that... :/ | 16:36 |
| @jim:acmegating.com | Clark: maybe the periodic cleanup is getting ahead of the deleted callback and so the deleted callback is repeating the work? | 16:37 |
| @clarkb:matrix.org | hrm ya self._register_lock. I was going to say a lock should solve this in this situation since it is all within a single object unlike what the idle check guards against. But not I'm confused | 16:38 |
| @jim:acmegating.com | (like the sequence is: 1) deleted callback is invoked; 2) cleanup starts and gets the lock; 3) cleanup re-registeres the deleted node and releases lock; 4) deleted callback resumes, obtains lock, and repeats the registration? | 16:38 |
| @jim:acmegating.com | (we should put some "end" debug lines in; that would help a lot here) | 16:39 |
| @clarkb:matrix.org | and may log when we get the lock | 16:40 |
| @clarkb:matrix.org | * and maybe log when we get the lock | 16:40 |
| @clarkb:matrix.org | should I work on that change? | 16:41 |
| @jim:acmegating.com | Clark: i think we have an end log entry for the cleanup -- "Cleaning up held nodes..." logs after the completion of cleanupLeakedResources | 16:45 |
| @clarkb:matrix.org | corvus: what is the trick for running multiple unittest jobs over and over in a change. inherit the actual job a number of times with new names? | 16:48 |
| @jim:acmegating.com | Clark: i think there's enough of a thread to pull on here that we might be able to identify/fix it. i'm happy to do that since i probably made the problem, but if you want to work on it i don't want to steal it from you. | 16:48 |
| @jim:acmegating.com | Clark: yes that's it | 16:48 |
| @clarkb:matrix.org | I've got a start going at least that will hopefully give us logging that we need | 16:49 |
| @clarkb:matrix.org | I can push that shortly and we can take it from there | 16:49 |
| -@gerrit:opendev.org- Clark Boylan proposed: [zuul/nodepool] 867031: Add extra debug logging to static driver https://review.opendev.org/c/zuul/nodepool/+/867031 | 16:51 | |
| -@gerrit:opendev.org- Zuul merged on behalf of Clark Boylan: [zuul/zuul-jobs] 866939: Move tox logs into env specific dirs https://review.opendev.org/c/zuul/zuul-jobs/+/866939 | 17:07 | |
| -@gerrit:opendev.org- Zuul merged on behalf of Clark Boylan: [zuul/zuul-jobs] 866935: Update zuul-jobs to handle tox3 and tox4 https://review.opendev.org/c/zuul/zuul-jobs/+/866935 | 17:23 | |
| -@gerrit:opendev.org- Sorin Sbârnea proposed wip: [zuul/zuul-jobs] 867037: Fixed schema errors reported by ansible-lint https://review.opendev.org/c/zuul/zuul-jobs/+/867037 | 17:50 | |
| @clarkb:matrix.org | corvus: the second pass of the extra logging change caught one (maybe two) https://storage.bhs.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_a3c/867031/1/check/nodepool-tox-py310-E/a3c287b/testr_results.html | 18:04 |
| @clarkb:matrix.org | now I need to make a call graph or something to see what is happening | 18:05 |
| @clarkb:matrix.org | https://paste.opendev.org/show/bOjNaNtmIqyxjZZ7vPRY/ now to make sense of that function call flow | 18:15 |
| @clarkb:matrix.org | based on that I suspect that the problem has to do with https://opendev.org/zuul/nodepool/src/branch/master/nodepool/driver/static/provider.py#L472-L477 since that occurs outside of the lock after thread B starts and while A is running | 18:16 |
| @jim:acmegating.com | Clark: there's a check in the syncNodeCount method that checks to see if we have a node registered, but the nodedeletednotification doesn't check that | 18:21 |
| @clarkb:matrix.org | ya I think they race around this line https://opendev.org/zuul/nodepool/src/branch/master/nodepool/driver/static/provider.py#L347-L350 | 18:22 |
| @jim:acmegating.com | so it's always going to register a new node even if the sync right before it already did that | 18:22 |
| -@gerrit:opendev.org- Sorin Sbârnea marked as active: [zuul/zuul-jobs] 867037: Fixed schema errors reported by ansible-lint https://review.opendev.org/c/zuul/zuul-jobs/+/867037 | 18:22 | |
| @clarkb:matrix.org | corvus: if the B thread goes first then A gets back a non Null value | 18:22 |
| @clarkb:matrix.org | and its fine | 18:22 |
| @jim:acmegating.com | Clark: so my inclination (still not 100% paged in here) is that nodeDeletedNotification needs to detect the case where it doesn't actually need to do anything. | 18:23 |
| @jim:acmegating.com | Clark: yep, it works correctly the other way | 18:23 |
| @jim:acmegating.com | (so if nodeDeletedNotification re-registers the node, then cleanup runs, cleanup knows it doesn't need to register) | 18:23 |
| @clarkb:matrix.org | I guess we just have to check the slots then? | 18:24 |
| @clarkb:matrix.org | and return if not None | 18:24 |
| @clarkb:matrix.org | inside the lock | 18:24 |
| @jim:acmegating.com | Clark: (btw i think 474-477 are okay outside the lock since that's just basically saying "is this node type even registered in the config file") | 18:24 |
| @jim:acmegating.com | > <@clarkb:matrix.org> inside the lock | 18:24 |
| yes to this series of msgs | ||
| @jim:acmegating.com | put another way, we shouldn't believe that the old node's slot is still empty (line 487) we should double check that | 18:25 |
| @jim:acmegating.com | Clark: i think the _getSlot call may be helpful | 18:27 |
| @clarkb:matrix.org | looking. I was trying to make sense of that enumerate call but maybe that is a short cut | 18:28 |
| @jim:acmegating.com | yeah i think so | 18:29 |
| @clarkb:matrix.org | hrm do we need to check all slots or just that the one that matches node is correct? | 18:29 |
| @clarkb:matrix.org | I guess in this case we are talking about something specific not the high level construct | 18:29 |
| @clarkb:matrix.org | since what is deleted is a single item | 18:29 |
| @jim:acmegating.com | given a nodeTuple, _node_slots gives us a list of Node objects (and the slot is just the index of each Node in that list -- that's why we get the node number from the enumerate) | 18:30 |
| @clarkb:matrix.org | `if self._getSlot(node): return` seem about right after `getRegisteredNodes()`? | 18:31 |
| @clarkb:matrix.org | oh index raises if it isn't there | 18:31 |
| @clarkb:matrix.org | so try except instead of if and return | 18:32 |
| @jim:acmegating.com | you know what, this may not be the best approach -- it's possible in the normal case that we've actually already removed the Node object from the node_slots list... | 18:32 |
| @jim:acmegating.com | i'm not sure about that, but it does strike me as something that would be reasonable to do.. so instead... | 18:33 |
| @jim:acmegating.com | maybe we should do the enumerate in order to find if there is a node in the slot specified by the deleted node | 18:33 |
| @jim:acmegating.com | (if there is a node in the slot, and it's not the same node that was deleted, then we know it was recreated. if there is no node in the slot, then i'm right about it having just been deleted before nodeDeletedNotification and it is correct to re-register it. if there is a node in the slot and it's the same as the one we're getting the notification about, then i'm wrong about that.) it's probably worth double checking what happens right before nodeDeletedNotification... | 18:35 |
| @clarkb:matrix.org | put another way if we _getSlot() and we get back None then we reregister. If we get back a node we return. If we have an exception we can probably return too since that indicates the node was properly removed form the config? | 18:38 |
| @jim:acmegating.com | okay i dug into that -- it's not the case that the node is deleted before nodeDeletedNotification... so scrap that last bit and let me type again... | 18:38 |
| @jim:acmegating.com | ignoring the cleanup handler for the moment, the normal workflow is: when a node is done, the zk node state is set to DELETED and then we call nodeDeletedNotifaction. that's what normally causes a new Node object to be put in the slot list (with the same slot number as before) | 18:40 |
| @jim:acmegating.com | now, bringing the cleanup handler back in -- what it does it that it sees there is a node in the slot list that's marked as deleted, so it doesn't count it; it discards it and replaces it. | 18:41 |
| @jim:acmegating.com | so back to what we should check for in nodeDeletedNotification: either we should find the exact node object that was passed in to the method in the expected slot (normal case), or we might find a different node object in that slot (that's the cleanup handler ran first case). i think we should only act in the first case and ignore any other case. | 18:43 |
| @clarkb:matrix.org | Got it so if node == self._node_slots[node_tuple][self._getSlot(node)]: then proceed else return? | 18:45 |
| @clarkb:matrix.org | we don't want to handle the None case we let the cleanup do that? | 18:45 |
| @clarkb:matrix.org | or do we also need to handle the None case? | 18:45 |
| @clarkb:matrix.org | fwiw I'm happy to review a change too if that is easier (I think you understand this much better than I do) | 18:45 |
| @jim:acmegating.com | > <@clarkb:matrix.org> Got it so if node == self._node_slots[node_tuple][self._getSlot(node)]: then proceed else return? | 18:46 |
| yeah, i think so. | ||
| @jim:acmegating.com | > <@clarkb:matrix.org> we don't want to handle the None case we let the cleanup do that? | 18:47 |
| i think so for this -- it seems to be doing a pretty good job of resyncing things. one way to look at it is that the delete notification is just a special case to speed up the process. | ||
| @clarkb:matrix.org | I see | 18:48 |
| @clarkb:matrix.org | ok I should be able to write that patch. I'll put it under my run all the unittests change so we can see how it does | 18:48 |
| @jim:acmegating.com | sounds good | 18:48 |
| -@gerrit:opendev.org- Clark Boylan proposed: | 18:59 | |
| - [zuul/nodepool] 867031: Add extra debug logging to static driver https://review.opendev.org/c/zuul/nodepool/+/867031 | ||
| - [zuul/nodepool] 867046: Correct race in static node reregistration https://review.opendev.org/c/zuul/nodepool/+/867046 | ||
| @clarkb:matrix.org | something like that? | 18:59 |
| @jim:acmegating.com | Clark: yeah, that lgtm | 19:04 |
| -@gerrit:opendev.org- Zuul merged on behalf of Jeremy Stanley https://matrix.to/#/@fungicide:matrix.org: [opendev/zone-zuul-ci.org] 866615: Restore the default TTL to lists https://review.opendev.org/c/opendev/zone-zuul-ci.org/+/866615 | 19:07 | |
| @q:fricklercloud.de | seeing static nodes mentioned reminds me that https://review.opendev.org/c/zuul/nodepool/+/863812 could still need a review or two | 19:40 |
| -@gerrit:opendev.org- Ian Wienand proposed: [zuul/nodepool] 866955: nodepool-functional-k8s: switch to microk8s https://review.opendev.org/c/zuul/nodepool/+/866955 | 20:33 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 867057: WIP: play with noxfile https://review.opendev.org/c/zuul/zuul/+/867057 | 21:13 | |
| @jim:acmegating.com | Clark: that's a super-minimal noxfile i was playing with yesterday | 21:14 |
| @jim:acmegating.com | probably needs to not have "/home/corvus" in there before it's generally useful ;) | 21:15 |
| @clarkb:matrix.org | ha, also how does it interpret -R py310 or whatever the incantation is for a specific python version? | 21:16 |
| @jim:acmegating.com | that bit where we get concurrency from python could probably be a lot easier... | 21:16 |
| @jim:acmegating.com | we can parameterize the "py3" session | 21:17 |
| @clarkb:matrix.org | ya you could lift it into the nox python I suspect | 21:17 |
| @jim:acmegating.com | https://nox.thea.codes/en/stable/tutorial.html#testing-against-different-and-multiple-pythons | 21:17 |
| @jim:acmegating.com | so then i think the cli would be "nox -R py3-3.11" but probably we want to rename it to "py3-3.11". i wonder if we can alias so we could do py311... i dunno | 21:18 |
| @jim:acmegating.com | * so then i think the cli would be "nox -R py3-3.11" but probably we want to rename it to "py-3.11". i wonder if we can alias so we could do py311... i dunno | 21:18 |
| @jim:acmegating.com | we could also not parameterize, so we could have a "py311" session and just have it call a common method. | 21:19 |
| @jim:acmegating.com | you know, function composition | 21:19 |
| @jim:acmegating.com | anyway, it looks like it might work for us, and i kinda think a python implementation of tools/pip.sh inside noxfile might be less hacky than what we have for tox | 21:20 |
| @clarkb:matrix.org | The biggest issue is likely just going to be getting familiar with the new tool | 21:21 |
| @clarkb:matrix.org | but we're facing similar problems with tox anyway | 21:21 |
| @jim:acmegating.com | yeah and for me, remembering to use "-R" to mean "reuse the env and install" | 21:21 |
| @michael_kelly_anet:matrix.org | corvus: I wanted to double check to ensure I'm not missing something, but afaict there's no support for configuring auth via the operator. Is this accurate? Or am I missing something? | 23:43 |
| @jim:acmegating.com | Michael Kelly: i believe that's the case yes; it probably should be modeled on the connection data structure | 23:48 |
| @michael_kelly_anet:matrix.org | Sounds good. I've got to circle back to some hacking (and updating/dealing with your review comments) next week, so I'll probably burn some cycles to hack that up as well when I refocus | 23:50 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!