Sign in

OOMKilled With A 1Gi Limit, But The Go Microservice 'only Uses' Half Of It????

Solveddockerkubernetesmemory3 replies · 290 views · started
Reply
Accepted solution by Chen N. — Set requests close to real usage and limits with honest headroom xD on Kubernetes, a pod with requests equal to limits gets Guaranteed QoS…
  1. Gabe R.
    @stackwrangler

    long time lurker first post lol migrated the Go microservice from a VM to containers... on teh VM it was fine for years, in a container with 1Gi:

    killed with 137 wtihin an hour of heavy traffic, same code, same traffic, what changed?

    on the VM, nothing ever complained lmao i suspect the runtime is behaving differently inside the container, but i dont know what to check :/ any advice on sizing a 1Gi cotnainer properly would be much appreciated. any ideas??

    5
  2. Chen N.
    @quartzlab-2c0
    Accepted solution

    Set requests close to real usage and limits with honest headroom xD on Kubernetes, a pod with requests equal to limits gets Guaranteed QoS and is teh last to be evicted under node pressure.

    3 4 helpful
  3. Otto K.
    @otto_mutex

    @stackwrangler did u ever fix it?... this, so much this ?

    1 13 helpful
  4. Gabe R.
    @stackwrangler
    in reply to @quartzlab-2c0

    @quartzlab-2c0 omg finally the nightly import loaded everything into memory at once. Switched to streaming and the OOMKills stopped at 1Gi.

    13 1 helpful

Reply to this topic

Sign in or create an account to reply.