-
Notifications
You must be signed in to change notification settings - Fork 209
Recommendations on how to configure PHP (Doctrine) to work with PGDog as a connection pooler? #892
-
We are hitting a lot of EOF detected, due to the fact that our pgdog pods get rescheduled by our autoscaler, and PHP is not able to recover from a lost connection.
Has anyone hit a similar problem? how would you solve that?
Beta Was this translation helpful? Give feedback.
All reactions
Replies: 2 comments
-
We've had some success with:
podAnnotations: cluster-autoscaler.kubernetes.io/safe-to-evict: "false"
Beta Was this translation helpful? Give feedback.
All reactions
-
@levkk thanks, using a dedicated node pool for pgdog or reducing the possibility a pgdog pod gets rescheduled, is something we thought too. But looks like we are loosing the advantages of running in kubernetes. At the moment we have 6 pod with a maximum of 1 unavailable at a time.
However I feel like the app should be able to handle broken connections and simply reconnect/retry. This is built-in with go, but we have legacy PHP apps that aren't as reliable. I wonder wether there is some configuration tuning or some middleware we could use in one of PHP or pgdog
Beta Was this translation helpful? Give feedback.