Open tanisdlj opened 3 years ago
Hi!
To be sure, don't you have messages like following, before one you posted?
Adding cluster manager for 'druid' with X dataCubes
Connected to cluster 'druid'
If that's the case, this is what happens:
Turnilo starts, reads your config file, but also tries to fetch metadata from druid cluster. This is safety check, so turnilo won't display non existing data cubes.
But if this metadata fetch takes a lot of time (timeout is hit), turnilo just returns settings from configuration. But fetch is handled in the background and after successful request, turnilo will get this metadata, and next client would get correct configuration.
I don't think that slows turnilo startup - you still need to get those metadata. If you wan't more predictable behaviour, you can increase this timeout - set pageMustLoadTimeout
in your config file to number of milliseconds.
There is also another possibility, but less likely. Maybe you have some health check service that hits turnilo during startup? It could also hit this timeout before even metadata fetch occurs. In that case, you just need to let turnilo start a little longer :)
To be honest, we would also benefit from "configuration as a source of truth" mode, where no metadata fetch is needed. But we don't have resources to implement this and still keep legacy behaviour :(
Thanks! I'll look into that.
By the way, the setting pageMustLoadTimeout
is not documented in turnilo/docs/configuration-cluster.md, it would be nice to add it there
By the way, the setting
pageMustLoadTimeout
is not documented in turnilo/docs/configuration-cluster.md, it would be nice to add it there
Good catch!
Hi! We're hitting a timeout when we start the service
It's not super critical, but makes the "full start" of turnilo a bit slower. Any clue why this might be happening? Is it a bug or a problem on my config?
Thanks a lot!