pgwatch3 as K8S operator

[ondar]

• Has anyone already implemented K8S helm charts for pgwatch3?
• Is it actually possible to have postgres and prometheus sinks at the same time? With multiwriter?
• Any idea about resource usage for 1 source and 2 sinks (timescale postgres + prometheus)?
• It is possible to disable UI to reduce resource usage and increase security?

[pashagolub]

• Has anyone already implemented K8S helm charts for pgwatch3?

There are some ongoing works. @Schmaetz any news on this?

• Is it actually possible to have postgres and prometheus sinks at the same time? With multiwriter?

Yes, this is by design. Yes, multiwriter is responsible for that. Just put several --sink= command-line parameters and pgwatch will catch those automatically

• Any idea about resource usage for 1 source and 2 sinks (timescale postgres + prometheus)?

Minimal. Depends on the intervals of metrics and scrapping. If compression is on for TimescaleDB and we assume
45M rows of real world stat_statements data (6db * 3mo), storage: 8 GB compressed.

But usually people have shorter retention periods, so it's even less.

• It is possible to disable UI to reduce resource usage and increase security?

Implemented in #576, just make sure --web-addr='' is empty

[ondar]

Thanks, I will update docs with this new info. And what about memory + cpu usage?

[pashagolub]

One needs to measure but I don't think it drastically increased comparing to previous version

[ondar]

Is it possible to somehow configure multiple sinks with PW_SINK env var, or you must use --sink cmd-line parameter? Same with --web-addr=''?

[pashagolub]

I didn't try to use several values with env var, need to check. if PW_WEB_ADDR is empty that means default value :8080 will be used, I suppose. What is the problem using hard coded command-line parameter to disable web ui?

[ondar]

Probably no problem, depends how those helm charts are created. I'm no expert at that, I'm just relaying questions from other people from our company. I will have more answers and maybe questions probably tomorrow.