Bot failed to label PR

[mscdex]

nodejs/node#9221

[phillipj]

Logs says it the bot never got a pull_request.opened event from github for that PR, only pull_request.edited which is strange. Never seen this happen before, might be good to take notice when this occurs and maybe ask github support if we see more of this.

[mscdex]

Saw another instance just a bit ago: nodejs/node#9332

[phillipj]

#9332 seems to be related to the fact that we only fetch the first 100 labels of the node repository.

Pulled from the logs:

07:27:13.409 DEBUG bot: Filled existing repo labels cache: confirmed-bug,duplicate,help wanted,invalid,question,wontfix,in progress,tls,crypto,buffer,http,https,assert,util,cluster,child_process,dgram,c++,doc,debugger,dns,domain,events,fs,module,net,os,querystring,readline,repl,stream,timers,url,vm,zlib,windows,install,semver-minor,semver-major,discuss,build,V8,test,freebsd,path,meta,string_decoder,benchmark,openssl,npm,land-on-master,good first contribution,feature request,process,console,libuv,mentor-available,os x,arm,tools,land-on-v3.x,tty,intl,memory,land-on-v0.12,land-on-v0.10,land-on-v4.x,post-mortem,lts-landed-on-v4.x,lts-watch-v4.x,smartos,lts,ctc-agenda,ctc-meeting,v0.10,v0.12,v4.x,master,v5.x,lts-agenda,lib / src,lts-watch-v0.10,lts-watch-v0.12,land-on-v5.x,unconfirmed,stalled,mips,dont-land-on-v5.x,security,dont-land-on-v4.x,performance,error-handling,repro-exists,async_wrap,known issue test,promises,docs-requested,blocked,dont-land-on-v6.x,v6.x (req_id=d5d46425-aff7-411c-892b-bbacb61bec85, pr=nodejs/node/#9332, action=pull_request.opened)

cares is not part of the cached list of labels :/

[mscdex]

Ah ok

[mscdex]

nodejs/node#9422 was not automatically tagged as V8. The number of labels is at 100 exactly, so that shouldn't have been a problem?

[phillipj]

You're right, the V8 label existed in the labels cache at that moment, so that's not the issue for that PR. Related logs:

19:55:17.778 DEBUG bot: Filled existing repo labels cache: confirmed-bug,duplicate,help wanted,invalid,question,wontfix,in progress,tls,crypto,buffer,http,https,assert,util,cluster,child_process,dgram,c++,doc,debugger,dns,domain,events,fs,module,net,os,querystring,readline,repl,stream,timers,url,vm,zlib,windows,install,semver-minor,semver-major,discuss,build,V8,test,freebsd,path,meta,string_decoder,benchmark,openssl,npm,good first contribution,feature request,process,console,libuv,mentor-available,os x,arm,tools,tty,intl,memory,land-on-v4.x,post-mortem,lts-watch-v4.x,smartos,lts,ctc-agenda,ctc-meeting,v0.10,v0.12,v4.x,lib / src,lts-watch-v0.12,stalled,mips,security,dont-land-on-v4.x,performance,repro-exists,async_wrap,known issue test,promises,docs-requested,blocked,dont-land-on-v6.x,v6.x,cares,http_parser,V8_inspector,addons,punycode,v7.x,async_hooks,dont-land-on-v7.x,errors,ctc-review,lts-watch-v6.x,land-on-v6.x,constants (req_id=98d7aef2-e978-4a06-a74d-04416e91927c, pr=nodejs/node/#9422, action=pull_request.opened)
19:55:17.778 DEBUG bot: Resolved labels: v6.x (req_id=98d7aef2-e978-4a06-a74d-04416e91927c, pr=nodejs/node/#9422, action=pull_request.opened)
19:55:17.778 DEBUG bot: Trying to add labels: v6.x (req_id=98d7aef2-e978-4a06-a74d-04416e91927c, pr=nodejs/node/#9422, action=pull_request.opened)
19:55:17.947  INFO bot: Added labels: v6.x (req_id=98d7aef2-e978-4a06-a74d-04416e91927c, pr=nodejs/node/#9422, action=pull_request.opened)

[mscdex]

I think it may be because of the change in capitalization (to differentiate between node v8 and V8 the Google project)? The directory name is lowercase (deps/v8) but the label is V8.

[phillipj]

Yupp, agreed. My best guess is the node-repo.js#itemsInCommon(), made a simple script to verify it:

function itemsInCommon (arr1, arr2) {
  return arr1.filter((item) => arr2.indexOf(item) !== -1)
}

const existing = ['V8']
const resolved = ['v8']

console.log(itemsInCommon(existing, resolved))
// []

[mscdex]

Perhaps we could just lowercase everything when comparing?

[mscdex]

New case: nodejs/node#9737

[phillipj]

Yet another case of not getting the pull_request.opened webhook event from GitHub :/ This is what's in our logs related to that PR.

10:25:02.306  INFO bot: Emitting GitHub event (req_id=a5192783-813c-40f0-b969-8322e98822b1, pr=nodejs/node/#9737, action=pull_request.edited)
10:25:02.775  INFO bot: Emitting GitHub event (req_id=8dd41e3b-738e-452c-9e35-d6290bf8f717, pr=nodejs/node/#9737, action=pull_request.edited)
10:36:53.804  INFO bot: Emitting GitHub event (req_id=ac3724ad-4b3d-42ad-a1ee-874e85e429be, pr=nodejs/node/#9737, action=pull_request.labeled)
10:36:53.823  INFO bot: Emitting GitHub event (req_id=10749934-da31-4c00-8610-92e01c90f004, pr=nodejs/node/#9737, action=pull_request.labeled)

[phillipj]

Just realised we've got some deployment issues to resolve before blaming GitHub for not sending us all the events we should have gotten.

1. All current bot logs gets wiped whenever the bot is deployed
2. We've got some downtime when deploying -- its a matter of seconds, tho zero downtime deploys would ease my mind a little.

Both might be the reason we either don't see events logged, cause the logs are simply not complete or an even was pushed by GitHub while the bot was deploying.

[mscdex]

@phillipj Was there a deploy recently? nodejs/node#10156 is another example of no labels.

[phillipj]

Nope, no recent deploys. Looking at the systemd service supervising the bot, it might restarted two days ago. That should not have affected the PR 10156 tho since that's one day old.

No traces of PR 10156 in the logs either 😞

[mscdex]

Interesting... this time the repo's webhook logs show the "opened" notification was sent successfully (instead of failed in previous PRs) for that PR and it shows Express returned with 200 OK.