We have not small database (~450Gb, ~30000 tables) and few slave servers for it:
- slave2 (scsi hdd with raid 10)
- slave3 (sas hdd wthout raid)
but after we doing “optimize” at master server replication at slave3 starts to lag:
- load on the disk becomes 100%
- the number of inserts falls
- lag continues to grow
It seems to be a hardware problem, but if we will get a new copy of Mysql datadir from slave2 (which have no lag) to slave3 - we will get an actual copy of database (when it will catch up). On graphs in cacti we can see that hardware can handle ~100 inserts per second, but when it’s 100% IO - there is only ~30 inserts per second.
What we alredy did:
- Move relay logs to another hdd
- disable journal at ext4 and tune it a little (mount opts - rw,noatime,nodiratime,data=writeback)
but no improvement…
What else to tune?