Cloud SQL Read Replica Legging behind master

[Rahul Goti]

It seems replication to read replica is not working properly with CloudSQL. Every day the replication delay graph spike is high and not draining. I'm using a master instance with 15GB memory/4 vCPUs while read replica is with 30GB memory/8 vCPUs. So I don't think there is a memory issue. Can you please help me out with this please?

[Olu]

This documentation[1] gives hints on the possible causes of replication lag, especially when they are consistently high. The replication delay often indicate that write workload may be too high for the replica to handle. I have seen situations in the past where the replication delay was a result of consistently handling workloads with CPU intensive queries (sorting, regular expressions, other complex functions). As explained in this documentation[2], CPU intensive queries often make the CloudSQL instance to be throttled.

For better understanding or the exact details of what may be happening on your CloudSQL Instance, I recommend you contact the GCP Support[3]. This way a GCP Support Engineer would be able to review the specific details on your CloudSQL with you.

[1]https://cloud.google.com/sql/docs/mysql/replication/tips#highlag

[2]https://cloud.google.com/sql/docs/mysql/diagnose-issues#tips

[3]https://cloud.google.com/support-hub