Closed fako closed 5 months ago
TimeBox 4h for research. Demo environment needed for testing.
There are a lot of file warnings. How are we going to fix this on the source side?
I get the feeling this is very much related to the write errors. Those seem to be failed Sentry logging attempts. How are we making this more sustainable? What is the frontend doing??
We are going to try to solve Pure image problems and expect these errors to go away or dramatically reduce. Renaming ticket to reflect this.
This is now deployed to dev for testing. We need to discuss how/when we want to deploy this to demo and beyond @peterdubbeldsurf @mlagerwerf
We should test a complete reset
We should also test streaming on frontend
I'm not sure we can fix these write errors on our side. The connections get disconnected at Pure's side at least sometimes. As is seen in this event: https://surf-eduservices.sentry.io/issues/5209011742/events/85bab88877fe4487ada6435e73d42619/?project=4505125577228288
Here's a screenshot that shows that these kinds of Pure disconnects happen together with dozens of write errors:
More example errors showing that Pure cuts off responses half way: https://surf-eduservices.sentry.io/issues/5313807052/?project=4505125577228288&query=is%3Aunresolved&referrer=issue-stream&statsPeriod=90d&stream_index=1
Playback van videos werkt, maar er is geen streaming. Dit zal geimplementeerd moeten worden in de frontend, maar de stream hopt dan via ten minste 4 "servers". De AWS balancer, de Nginx frontend, de UWSGI backend en Pure. Waarschijnlijk heeft Pure ook nog balancers en verschillende middleware. Dit is in ieder geval het geval bij Hanze, die een eigen middleware tussen Pure en de Harvester heeft gezet.
Hanze voorbeeld: https://client.publinova.acc.surf.zooma.cloud/product/de-gebiedscooperatie-als-model-voor-leven-lang-ontwikkelen
@peterdubbeldsurf
Problem occurs when "proxy files" is enabled for HvA