MySQL Forums
Forum List  »  InnoDB

Re: Table freeze when size exceeds innodb buffer pool size
Posted by: Peter Brawley
Date: September 28, 2017 04:54PM

Select_scan / Uptime_since_flush_status --> table scans/sec.

Handler_read_rnd_next / Uptime_since_flush_status --> such row requests/sec.

> Is quite hard to upgrade cause its a live system

Um, if we all followed that rule, no system in use would ever be upgraded.

To do it in minimal downtime you need to develop & debug a failsafe upgrade script, then optimise that till the donwtime is satisfactorily low, then do it at an offpeak time. So you need three systems: production, backup, test/devel.

> InnoDB log files ... How you derive on this? What size is normally best

innodb_os_log_written / Uptime_since_flush_status / innodb_log_files_in_group, rounded uo to a 32MB boundary. What's best is often about 30 mins of writes.

> InnoDB buffer pool appears to be running 50% free ... if I shrink will I need to expand later when the table size grows?

Yex it can and should be dynamic if the system is growing. Check every month or so.

Options: ReplyQuote




Sorry, only registered users may post in this forum.

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.