Hi,
In our company we user Postgres v 9.4, with streaming replication (master + one replica).
We use WAL-E for backups (WAL files continuously archiving to S3), and use wal-e backup-push every night.
Unfortunately, I see performance degradation while backup-push running, so I use "--cluster-read-rate-limit" option.
For now, backup starting at 05:00 and ends at 12:00 (7 hours). And this time getting bigger and bigger every day. I am not satisfied with this solution.
We have one replica node. Is there any options to use replica as a source for backups? Or some other scenarios?
Also, our app running in OpenStack cloud and we can easily do snapshots for our disks. Maybe we should use this feature to solve very long backup time?
Thank you!
--