Bug #29388 innodb redo log size rule?
Submitted: 27 Jun 2007 13:53 Modified: 11 Jul 2007 15:00
Reporter: Kristian Koehntopp Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Enterprise Monitor: Advisors/Rules Severity:S3 (Non-critical)
Version:1.1.0 OS:Any
Assigned to: Andy Bang CPU Architecture:Any

[27 Jun 2007 13:53] Kristian Koehntopp 
Description:
InnoDB Redo Logs Not Sized Correctly  (v 1.3 *)

Thresholds
 Critical Alert = 20
 Warning Alert = 30
 Info Alert = 40 

Expression
(%have_innodb% == "YES") && ((100*(%innodb_log_file_size% / %innodb_buffer_pool_size%)) < THRESHOLD) && (%innodb_buffer_pool_size% < 2147483648)

1. Does not take log files in group into account.
2. What if I have more than 2G of pool?
3. Isn't the redo log size calculation more complicated? (Log waits, and innodb_max_dirty_pct both forcing buffer pool flushes).

How to repeat:
Look at the rule, do the math for a 16G pool server.

Suggested fix:
PeterZ recommendation: Make the redo log 128-512M in size (%innodb_log_file_size% * %innodb_log_files_in_group%) between 128M and 512M, then check for log waits. If you see them, find out why they happen. That may be complicated (sustained writes -> more spindles, very high peaks -> maybe larger redo logs, but caveat recovery times)
[10 Jul 2007 21:21] Andy Bang 
Committed revision 6546.
[11 Jul 2007 10:32] Mark Leith 
Verified fixed in 1.2.0.6550
[11 Jul 2007 10:37] Mark Leith 
one for the changelog..
[11 Jul 2007 15:00] Peter Lavin 
Added to the changelog for version 1.2.
[11 Jul 2007 15:00] Peter Lavin 
Added to the changelog for version 1.2.