Bug #21956 Test failure: "query cache not invalidated"
Submitted: 31 Aug 2006 20:27 Modified: 22 Oct 2009 9:02
Reporter: Joerg Bruehe Email Updates:
Status: No Feedback Impact on me:
None 
Category:Tests: Cluster Severity:S3 (Non-critical)
Version:5.0.20a and up OS:Mac OS X (OS-X 10.4 PPC 64bit)
Assigned to: Assigned Account CPU Architecture:Any

[31 Aug 2006 20:27] Joerg Bruehe
Description:
This test failure occurs since (at least) 5.0.20a,
it is specific to this single platform.

It happens in "cluster" only, as currently this is the only product there including NDB (which is currently excluded from "max", except for my test today).

Symptom:

=====
ndb_cache_multi2               [ fail ]

Errors are (from /.../mysql-max-5.0.25-pre-osx10.4-powerpc-64bit/mysql-test/var/log/mysqltest-time) :
mysqltest: At line 74: query 'The query_cache thread failed to invalidate query_cache in 10 seconds' failed: 1064: You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near 'The query_cache thread failed to invalidate query_cache in 10 seconds' at line 1
=====

The test contains a loop of 20 turns, each one with a "sleep 0.5", this is how those 10 seconds are to be determined.

I am not sure whether this is an accurate method, IMO it depends very strongly on machine load etc (on this machine, we always run a 32 and a 64 bit build in parallel)

How to repeat:
Build including NDB, run the tests.

Suggested fix:
I am not sure whether this is an accurate method:

1) IMO it depends very strongly on machine load etc (on this machine, we always run a 32 and a 64 bit build in parallel).

2) Is "sleep 0.5" really implemented correctly ?

3) What is the base for the claim that after 10 seconds the cache must have been invalidated ?

So IMO this may well be a design problem of the test, need not indicate a server problem.
[1 Feb 2008 16:16] Joerg Bruehe
Still occurs in the release builds,
just now in 5.1.23-rc
[22 Oct 2009 23:00] Bugs System
No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".