Re: Daisy Chained Replication Lag
A Slave getting behinder and behinder --
* The Master can can perform transactions in parallel; but the slaves (by default) cannot...
* Look into multi-threaded Slave thread. If you are using Oracle MySQL, then you can split the processing on the slave between databases. If you have only one database, no help here. OTOH, Maria DB has better splitting. I see you have slave_parallel_workers=20; are they useful?
* Optimize the queries on the Master (after all, they have to go through to the slaves). On the Master, turn on the Slowlog with long_quer_time=1, wait a day, then run pt-query-digest to see what the worst hogs. Ignore SELECTs, then ponder what can be done to speed up the top could of writes. Post them, together with SHOW CREATE TABLE, on the performance forum for help.
* What binlog_format are you using? For mostly 1-row writes, ROW is less burden on Slaves. For massive UPDATE/DELETEs, STATEMENT saves a lot on network bandwidth.
* Massive DELETEs -- Can they be turned into DROP PARTITION?
* What version are you using; newer versions _may_ have improvements that would be useful.
Subject
Views
Written By
Posted
2230
June 23, 2015 10:43AM
Re: Daisy Chained Replication Lag
1207
June 30, 2015 05:28PM
1088
July 06, 2015 06:49AM
937
July 08, 2015 07:38PM
Sorry, you can't reply to this topic. It has been closed.
Content reproduced on this site is the property of the respective copyright holders.
It is not reviewed in advance by Oracle and does not necessarily represent the opinion
of Oracle or any other party.