Incident retro - home page and what's on not available
PreviousIncident retro - home page and stories page not availableNextIncident retro - requests not showing in account
Last updated
Last updated
Incident from: 2022-02-24
Incident until: 2022-02-24
Retro held: 2022-03-01
23 February 2022 PR merged Remove place from event (Prismic model and app types)
The model change happened at about 17:00 on the 23rd Feb
24 February 2022
See https://wellcome.slack.com/archives/C01FBFSDLUA/p1645692506693739
08.47 Alert down for:
Experience: Content: Homepage (cached)
Experience: Content: Homepage (origin)
Experience: Content: What’s on (cached)
Experience: Content: What’s on (origin)
08.48 AC I reckon this is probably @dmc’s change to the Prismic model – the deployed code is still looking for “place”, but it can’t find it because we’ve removed it And deploying the front-end is blocked because it’s broken in a different way I believe this fix will get the deployment unblocked: https://github.com/wellcomecollection/wellcomecollection.org/pull/7704
08.50 AC I’m going to merge #7704 by fiat and we can review it later, to unblock the deployment and get the homepage working again Specifically: we check a set of URLs for 200'ing as part of the deployment process, which you can see here: https://buildkite.com/wellcomecollection/experience-deployment/builds/657#9cf35e09-cf96-467b-becf-2fdaf1125598 One of those URLs is 500'ing, which means we shouldn’t be allowed to deploy… but it’s not as bad as the prod front-end, which is totally broken
08.52 Statuspage update created for home page and events not available
09.06 AC: still broken https://buildkite.com/wellcomecollection/experience-deployment/builds/658#7e887723-5ecf-4ac7-8a1b-8bba6add23c4
DMc but I’m wondering if putting place back into the Prismic model would make sense
09.10 AC depends how long it would take to propagate
I suspect the issue is that Prismic has only just propagated the model change you applied yesterday
09.25 PB Is it relevant that place is mentioned in this graphQuery?
09.26 DMc I think this is another possible issue https://github.com/wellcomecollection/wellcomecollection.org/pull/7705/files#diff-0d61a4c2769aa8f2991eec03d5ea6[…]ff8cd8a6b02a6def9ff904185f3befR221
I had thought that putting place back on the Prismic model would be fairly instant, but perhaps Alex was right about propagation in any event, this change brings the site back up locally for me PB’s review of the above isn’t powerful enough
09.32 DMc couldn’t add PB to https://github.com/orgs/wellcomecollection/teams/js-ts-reviewers
09.37 DMc couldn’t force the PR or push to main
09.38 Then I suppose we must wait for @Gareth Estchild or for the interviews to finish
09.40 PB asked JT to add him to https://github.com/orgs/wellcomecollection/teams/js-ts-reviewers / done at 09.44
09.49 PB While we wait for that to go through - Is the underlying cause of this down to the delay in Prismic propagation? i.e. that this all seemed fine and smoke tested perfectly because Prismic was still using the old model, so the queries using the deleted fields completed successfully, but then once it was deployed, and Prismic updated, it all went a bit wrong
10.07 Recovery for:
Experience: Content: What’s on (origin)
Experience: Content: What’s on (cached)
Experience: Content: Homepage (cached)
Experience: Content: Homepage (origin)
10.11 Statuspage update: resolved
Is the underlying cause of this down to the delay in Prismic propagation? i.e. that this all seemed fine and smoke tested perfectly because Prismic was still using the old model, so the queries using the deleted fields completed successfully, but then once it was deployed, and Prismic updated, it all went a bit wrong
Reading between the lines in the Prismic docs and an old Prismic support thread, coupled with the timing of content being published and the site going down (see below). It looks like the model updates are only reflected in the API response once a piece of content has been published.
N.B. The model change happened at about 17:00 on the 23rd Feb
We need to remember to publish something, when we change the model.
A quicker fix in this type of scenario therefore, would be to change the model back to its previous state and publish something, to see those changes in the API.
PB couldn’t approve DMcs PR
DM couldn’t add PB to https://github.com/orgs/wellcomecollection/teams/js-ts-reviewers
DMc couldn’t force the PR
Alex
All developers should be able to merge pull requests
Add dev permissions to onboarding/leaving checklist
Investigate splitting out the experience build
Make the monitoring lambdas vend a prefilled link to the logging cluster
David
Gareth
09.23 DMc PR:
Make the e2e tests go faster: If they were faster, we'd be able to recover from broken builds a lot faster.
Think about having a staging version of Prismic () and talk to other developers about what has been found out
Document how to re-add deleted fields in Prismic, and update a piece of content arbitrarily.
Add a message to the diff tools script re updating content after field deletion from the model.
Make a ticket: Investigate all uses of graph query, and make sure we’re only using it where needed