Bug #35503 falcon_bug_34351_C.test times out on Pushbuild
Submitted: 22 Mar 2008 19:19 Modified: 13 Dec 2008 7:22
Reporter: Hakan Küçükyılmaz Email Updates:
Status: Closed Impact on me:
Category:MySQL Server: Falcon storage engine Severity:S3 (Non-critical)
Version:6.0, 6.0-falcon-team OS:Any
Assigned to: John Embretsen CPU Architecture:Any
Tags: F_TEST, pb2, pushbuild, test failure

[22 Mar 2008 19:19] Hakan Küçükyılmaz
falcon_bug_34351_C.test times out on Pushbuild.

How to repeat:
To reproduce run falcon_bug_34351_C.test several times until you see a deadlock which leads to test timeout.
[24 Mar 2008 22:26] Sveta Smirnova
Thank you for the report.

Verified as described.
[28 Sep 2008 15:57] Hakan Küçükyılmaz
I don't see a deadlock anymore but a steady memory usage. See Bug#39711 for details.
[28 Sep 2008 20:54] Hakan Küçükyılmaz
See memory leak issue in Bug#39711 instead.
[11 Nov 2008 15:29] John Embretsen
Re-opening this bug, as the test still times out, but (it seems) only on Freebsd7 platforms in Pushbuild 2 (PB2). These platforms are not used in Pushbuild 1, nor in team-level branches hooked to PB2.

Fails in mysql-6.0 branch, example:

(test-max-freebsd7-x86, test-max-freebsd7-x86_64)
[11 Nov 2008 16:51] Kevin Lewis
John,  Both 34351_A and 34351_C could probably be cut in half, from a while loop of 10000 to 5000.  If you want, you could make -big versions as well with a loop of maybe 20000.  These tests used to get update conflicts given the right timing.
[20 Nov 2008 15:58] Bugs System
A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:


2916 John H. Embretsen	2008-11-20
      Fix for Bug#35503 - falcon_bug_34351_C.test times out on Pushbuild.
      Reducing loop count inside procedure from 10k to 5k in both falcon_bug_34351_A
      and falcon_bug_34251_C, as per Kevin Lewis' recommendations.
      Test case running time is effectively cut in half by doing this.
      Also adding new -big versions of those tests, running 20k iterations, in case 
      the bug is not triggered by 5k iterations in the non-big version. 
      These tests need to be run using the --big-test option of mysql-test-run.pl.
      On slower machines one might have to increase the test case timeout as well,
      e.g. --testcase-timeout=30 for a timeout of 30 minutes.
[20 Nov 2008 16:08] John Embretsen
Running times for all affected test cases on a test host running RHEL 5, 4x2800 MHz CPU, background load, given the patch described at http://lists.mysql.com/commits/59411:

TEST                           RESULT         TIME (ms)  TIME (approx. min)

falcon.falcon_bug_34351_A-big  [ pass ]         506750   8.5
falcon.falcon_bug_34351_A      [ pass ]         125638   2
falcon.falcon_bug_34351_C-big  [ pass ]        1012260   17
falcon.falcon_bug_34351_C      [ pass ]         254212   4+
[20 Nov 2008 16:45] John Embretsen
If test running time increases, it could possibly be related to bug#36700 (Running falcon_bug_28048 shows increasing memory usage and run time)...
[25 Nov 2008 10:09] John Embretsen
Actually, running time for non-big tests (34351 A and C) is less than 30 seconds on the linux host mentioned above when there is no background load. Should be relatively safe for current Pushbuild hosts.
[25 Nov 2008 19:16] John Embretsen
Note on the previous note: Smallish running times (< half minute) were measured using the --mem mysqltest option. No --mem still results in about the same results as reported earlier.

Patch pushed to mysql-6.0-falcon-team branch on 2008-11-25 13:44:28 GMT.
[1 Dec 2008 14:08] John Embretsen
See Bug#39711 (Running falcon_bug_34351_C shows increasing memory usage) for further investigations regarding running time and memory usage related to falcon_bug_34351_C.test.
[11 Dec 2008 14:20] Bugs System
Pushed into 6.0.9-alpha  (revid:john.embretsen@sun.com-20081120155823-v4dqfzqlkjqa6y5r) (version source revid:hky@sun.com-20081127084516-nbu7693932vcz2st) (pib:5)
[13 Dec 2008 7:22] MC Brown
No documentation needed.