For some reason, we want have full and incremental backup to support point in time restore, so we’ve follow the mongodb official doc to do backup for sharded cluster , but when we lock a secondary of config server to do mongodump, it would cause query timeout in production, something related to “uptime reporter” which is refreshing config setting with a majority read , but isn’t fsyncLock only lock writes ? i don’t get it , can anyone expain why it will cause production query timeout , is it a block method that would block all the query to the cluster ? what if we does not lock config but just lock shard secondary for backup ?
Plus, no exact point in time backup needed, just almost point in time backup is enough for us.
1 Like
fsynclock a single secondary shouldn’t be a problem as long as you have other 2 functioning nodes… in any case I suggest you give a try to Percona Backup for MongoDB which will automatically backup the sharded cluster for you
1 Like