Bug #30599 Thresholds too high for "Lock contention excessive" rule
Submitted: 23 Aug 2007 15:23 Modified: 2 Sep 2007 3:10
Reporter: Andy Bang Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Enterprise Monitor: Advisors/Rules Severity:S3 (Non-critical)
Version: OS:Any
Assigned to: Andy Bang CPU Architecture:Any
Tags: mer 120

[23 Aug 2007 15:23] Andy Bang
Description:
[08:10] <Isotopp> dev-merlin: heatchart rule "Lock contention excessive" has thresholds of 30, 60, 95%.
[08:11] <Isotopp> dev-merlin: why are they so high?
[08:11] <Isotopp> usually, 1% is harmful, and 3% is deadly.
[08:11] <Isotopp> (rule has version 1.4)
[08:12] <andy|away> Isotopp: the original thresholds were set in consultation with mark leith
[08:12] <andy|away> in going through lots and lots of rules for the first release, it's easy to miss something
[08:12] <andy|away> Isotopp: sounds like we should fix this one
[08:13] <Isotopp> maybe it is one of these delta things again
[08:13] <Isotopp> i usually use show global status values table_locks_immediate and _waited
[08:13] <Isotopp> and look at 1% and 3%.
[08:13] <Isotopp> still, booking does have issues, and the rule has never even fired an info alert, ever.
[08:14] <Isotopp> changed thresholds to 1, 3 and 5 now, which is extreme to the other side. will see tomorrow what comes up
[08:15] <andy|away> Isotopp: thanks.  please let us know what you find
[08:15] <Isotopp> sure

How to repeat:
See Description.
[23 Aug 2007 15:44] Andy Bang
[08:34] <Isotopp> getting info alerts (1%) for "lock connections" :) now.
[08:34] <Isotopp> this is reasonable
[08:34] <Isotopp> and even for the right machines.
[08:34] <Isotopp> i recommend 1, 2 and 3 percent. or 1, 3 and 5 (more conservative)
[08:35] <Isotopp> and gone till tomorrow
[24 Aug 2007 16:38] Andy Bang
Fixed in revision 7280.  Thresholds changed to 1%, 3%, and 5%.
[24 Aug 2007 16:43] Andy Bang
Andy Bang wrote:
> Mark,
>
> What do you think of Isotopp's proposal for lock contention thresholds (see
> below)?  Note that the expression is:
>
>     <expression><![CDATA[(%Uptime% > 10800) && (((%Table_locks_waited% /
> (%Table_locks_immediate% + %Table_locks_waited%)) * 100) >
> THRESHOLD)]]></expression>
> ...

Sure!

1, 3 and 5 would be my choice.

Cheers,

Mark
[2 Sep 2007 3:10] Andrew Cwik
fixed in 1.2.0.7406.