cryptnono: declare cpu/mem requests and limits #2451
Conversation
| # resources configuration configure both containers. | ||
| # | ||
| # PromQL queries for CPU and memory use: | ||
| # - CPU: sum(rate(container_cpu_usage_seconds_total{container="kube-trace-init", namespace="support"}[5m])) by (pod) |
There was a problem hiding this comment.
This query is just looking for the initcontainer, and not the actively running container (which is called 'trace' https://github.com/yuvipanda/cryptnono/blob/8367fa57835d912524c34fa06698bde6dbbae158/cryptnono/templates/daemonset.yaml#L42). Should account for that too, no, as it is the one that's constantly on?
There was a problem hiding this comment.
Yepp I checked both containers, the commented commands reflects only on the initcontainer and not both. The PR description includes some screenshots of the findings.
I've tried to come up with a request/limit etc based on both containers are started in sequence using the same resource allocation.
There was a problem hiding this comment.
I figure the queries can get the reader started and together with the text describing that there are two different containers can inspect both.
We've seen the cryptnono pod be evicted because it has no resource requests/limits when what needs to be evicted are other pods to resolve the situation. This is not a big deal, but we can set some resource requests and limits to avoid the situation - especially when cryptnono has started up fully and no longer will use much memory and cpu.
For details about requests/limits, see inline comments about cryptnono having an init container and a main container, where the main container is trivial compared to the init container.
Typically, both the init container and the main container are active only for ~4 minutes, following a period of ~0 CPU use.
CPU (leap, 30 last days)
Main container
Init container
Memory (leap, 30 last days)
Main container
Init container
