1. does it happen with all images or a specific one?
2. can you reproduce it constantly or is it intermittent?

I've seen similar reports before, but I honestly don't know how to reproduce. I've hosted Fider for 5 years and I never seen that issue.

Unfortunately it's quite random :( I have an automatic restart when it freezes and I got more than 50 restarts in a few months. I don't think my users are doing anything special, it's mostly PNG screenshots as attachments.

Hello, I'm having the same problem. It seems to be random and always when displaying images.
I added a LivenessProbe (k8s) so that the application is automatically restarted when this problem happens.

I can reproduce sometimes using a specific topic with some PNG attachments and then refreshing the entire page with the cache disabled.
Although even when using the LOG_LEVEL=debug, I don't feel that I have any useful logs:

  |   | INFO [2022-05-05T21:56:10Z] [WEB] GET http://fider.x.y.z/ finished with 0 in 1993ms (begin_error)
  |   | INFO [2022-05-05T21:56:08Z] [WEB] GET http://fider.x.y.z/ started
  |   | INFO [2022-05-05T21:56:05Z] [WEB] GET http://fider.x.y.z/ finished with 0 in 1995ms (begin_error)
  |   | INFO [2022-05-05T21:56:03Z] [WEB] GET http://fider.x.y.z/ started
  |   | INFO [2022-05-05T21:56:00Z] [WEB] GET http://fider.x.y.z/ finished with 0 in 2000ms (begin_error)
  |   | INFO [2022-05-05T21:55:58Z] [WEB] GET http://fider.x.y.z/ started
  |   | INFO [2022-05-05T21:55:55Z] [WEB] GET http://fider.x.y.z/ finished with 0 in 1999ms (begin_error)
  |   | INFO [2022-05-05T21:55:53Z] [WEB] GET http://fider.x.y.z/ started
  |   | INFO [2022-05-05T21:55:50Z] [WEB] GET http://fider.x.y.z finished with 0 in 1986ms (begin_error)
  |   | INFO [2022-05-05T21:55:48Z] [WEB] GET http://fider.x.y.z/ started
  |   | INFO [2022-05-05T21:55:45Z] [WEB] GET http://fider.x.y.z/ finished with 0 in 1999ms (begin_error)
  |   | INFO [2022-05-05T21:55:43Z] [WEB] GET http://fider.x.y.z/ started
  |   | INFO [2022-05-05T21:55:43Z] [WEB] GET http://fider.x.y.z/ finished with 0 in 4999ms (begin_error)
  |   | INFO [2022-05-05T21:55:40Z] [WEB] GET http://fider.x.y.z/ finished with 0 in 2000ms (begin_error)

When that happens, could you please check if there are any open transactions on Postgres? I believe this could be connection exhaustion, but it'd be useful to have more data.

I'm almost sure it's not connection exhaustion because I have another app using the same database (so a shared max connections) and it doesn't have connection pool issues.

Fider has its own connection pool, that's the one I think could be exhausted. You could also try increasing the pool to see if this reduces how often it happens

I suspect this could be the reason because I only use S3 to store attachments.

@goenning I managed to catch the pg_stat_activity table when the issue happens and I have 4 opened connections in the "idle in transaction" state and query SELECT provider_uid, provider FROM user_providers WHERE user_id = $1. I didn't update the DATABASE_MAX_OPEN_CONNS env var yet so your idea seems to be proven.

Fider has its own connection pool, that's the one I think could be exhausted. You could also try increasing the pool to see if this reduces how often it happens

fider/app/pkg/env/env.go

Line 63 in d107cbb

MaxOpenConns int `env:"DATABASE_MAX_OPEN_CONNS,default=4,strict"`

I suspect this could be the reason because I only use S3 to store attachments.

Nice one!
I've always been able to reproduce this issue on a page containing 14 attachments and refreshing with cache disabled.
When I increased the DATABASE_MAX_OPEN_CONNS env the problem disappeared. But I will continue to monitor.

Thanks for now!