[MDEV-7189] main.processlist fails sporadically in buildbot - Jira

Details

Type: Bug
Status: Closed (View Workflow)
Priority: Major
Resolution: Fixed
Affects Version/s: 5.5(EOL), 10.0(EOL), 10.1(EOL)
Fix Version/s: 5.5.42, 10.0.16, 10.1.3
Component/s: Tests
Labels:
- buildbot
- tests

Description

http://buildbot.askmonty.org/buildbot/builders/win32-debug2/builds/6568/steps/test/logs/stdio

main.processlist                         [ fail ]

        Test ended at 2014-11-24 05:14:33

CURRENT_TEST: main.processlist

--- C:/bzr/bb-win32/win32-debug2/build/mysql-test/r/processlist.result	2014-10-29 01:18:28.187152000 -0700

+++ C:\bzr\bb-win32\win32-debug2\build\mysql-test\r\processlist.reject	2014-11-24 05:14:30.816286800 -0800

@@ -12,5 +12,5 @@

 select command, time < 5 from information_schema.processlist where id != connection_id();

 command	time < 5

-Sleep	1

+Query	0

 set debug_sync='reset';

mysqltest: Result content mismatch

Variation with ps-protocol:
http://buildbot.askmonty.org/buildbot/builders/kvm-dgcov-jaunty-i386/builds/4530/steps/test_3/logs/stdio

main.processlist                         w2 [ fail ]

        Test ended at 2014-12-22 01:07:55

CURRENT_TEST: main.processlist

--- /home/buildbot/build/mysql-test/r/processlist.result	2014-12-21 20:37:16.794532000 +0200

+++ /home/buildbot/build/mysql-test/r/processlist.reject	2014-12-22 01:07:55.454312832 +0200

@@ -16,5 +16,5 @@

 SET DEBUG_SYNC = 'now WAIT_FOR query_done';

 select command, time < 5 from information_schema.processlist where id != connection_id();

 command	time < 5

-Sleep	1

+Execute	0

 set debug_sync='reset';

mysqltest: Result length mismatch

Another variation:
http://buildbot.askmonty.org/buildbot/builders/work-amd64-valgrind/builds/6424

main.processlist                         w1 [ fail ]

        Test ended at 2014-12-02 05:34:58

CURRENT_TEST: main.processlist

--- /var/lib/buildbot/maria-slave/work-opensuse-amd64/build/mysql-test/r/processlist.result	2014-12-01 23:07:01.000000000 +0200

+++ /var/lib/buildbot/maria-slave/work-opensuse-amd64/build/mysql-test/r/processlist.reject	2014-12-02 05:34:58.000000000 +0200

@@ -6,7 +6,6 @@

 SET DEBUG_SYNC = 'now SIGNAL fill_schema_proceed';

 INFO	TIME	TIME_MS

-NULL	0	0.000

 select sleep(5);

 sleep(5)

mysqltest: Result length mismatch

Attachments

Issue Links

blocks

MDEV-7069 Fix buildbot failures in main server trees

Stalled

MDEV-7172 Fix buildbot failures in 10.1 tree

Closed

Activity

Ascending order - Click to sort in descending order

Elena Stepanova created issue - 2014-11-24 16:08

Elena Stepanova made changes - 2014-11-24 16:08

Field	Original Value	New Value
Link		This issue blocks ~~MDEV-7172~~ [ ~~MDEV-7172~~ ]

Elena Stepanova made changes - 2014-11-24 16:30

Affects Version/s		10.0 [ 16000 ]
Affects Version/s		5.5 [ 15800 ]

Elena Stepanova made changes - 2014-11-24 16:31

Fix Version/s		10.0 [ 16000 ]
Fix Version/s		5.5 [ 15800 ]

Elena Stepanova made changes - 2014-11-24 16:31

Link

This issue blocks MDEV-7069 [ MDEV-7069 ]

Elena Stepanova added a comment - 2014-11-24 16:48

Updated affected versions according to the cross-reference report

Elena Stepanova added a comment - 2014-11-24 16:48 Updated affected versions according to the cross-reference report

Elena Stepanova made changes - 2014-12-23 16:06

Description

http://buildbot.askmonty.org/buildbot/builders/win32-debug2/builds/6568/steps/test/logs/stdio

{noformat}
main.processlist [ fail ]
        Test ended at 2014-11-24 05:14:33

CURRENT_TEST: main.processlist
--- C:/bzr/bb-win32/win32-debug2/build/mysql-test/r/processlist.result 2014-10-29 01:18:28.187152000 -0700
+++ C:\bzr\bb-win32\win32-debug2\build\mysql-test\r\processlist.reject 2014-11-24 05:14:30.816286800 -0800
@@ -12,5 +12,5 @@
0
select command, time < 5 from information_schema.processlist where id != connection_id();
command time < 5
-Sleep 1
+Query 0
set debug_sync='reset';

mysqltest: Result content mismatch
{noformat}

Variation with ps-protocol:
http://buildbot.askmonty.org/buildbot/builders/kvm-dgcov-jaunty-i386/builds/4530/steps/test_3/logs/stdio
{noformat}
main.processlist w2 [ fail ]
        Test ended at 2014-12-22 01:07:55

CURRENT_TEST: main.processlist
--- /home/buildbot/build/mysql-test/r/processlist.result 2014-12-21 20:37:16.794532000 +0200
+++ /home/buildbot/build/mysql-test/r/processlist.reject 2014-12-22 01:07:55.454312832 +0200
@@ -16,5 +16,5 @@
SET DEBUG_SYNC = 'now WAIT_FOR query_done';
select command, time < 5 from information_schema.processlist where id != connection_id();
command time < 5
-Sleep 1
+Execute 0
set debug_sync='reset';

mysqltest: Result length mismatch
{noformat}

Another variation:
http://buildbot.askmonty.org/buildbot/builders/work-amd64-valgrind/builds/6424
{noformat}
main.processlist w1 [ fail ]
        Test ended at 2014-12-02 05:34:58

CURRENT_TEST: main.processlist
--- /var/lib/buildbot/maria-slave/work-opensuse-amd64/build/mysql-test/r/processlist.result 2014-12-01 23:07:01.000000000 +0200
+++ /var/lib/buildbot/maria-slave/work-opensuse-amd64/build/mysql-test/r/processlist.reject 2014-12-02 05:34:58.000000000 +0200
@@ -6,7 +6,6 @@
1
SET DEBUG_SYNC = 'now SIGNAL fill_schema_proceed';
INFO TIME TIME_MS
-NULL 0 0.000
select sleep(5);
sleep(5)
0

mysqltest: Result length mismatch
{noformat}

Elena Stepanova added a comment - 2014-12-23 17:18

knielsen,

I see you already made some attempts to fix the test, but apparently there are still some flaws.

The first two variations in the description seem to be the same problem that you were fixing; but the 3rd one is puzzling, looking at the test I can't see how it can possibly happen.

Elena Stepanova added a comment - 2014-12-23 17:18 knielsen , I see you already made some attempts to fix the test, but apparently there are still some flaws. The first two variations in the description seem to be the same problem that you were fixing; but the 3rd one is puzzling, looking at the test I can't see how it can possibly happen.

Elena Stepanova made changes - 2014-12-23 17:18

Assignee

Elena Stepanova [ elenst ]

Kristian Nielsen [ knielsen ]

Kristian Nielsen made changes - 2015-01-06 12:36

Status

Open [ 1 ]

In Progress [ 3 ]

Kristian Nielsen added a comment - 2015-01-06 12:54

I think the third one is already fixed, in revid:knielsen@knielsen-hq.org-20141031114817-art3oxugmfgfd6ca

This fix appeared in 5.5 around 2014-10-31 and in 10.0 around 2014-11-03, and there are no failures of that kind seen in cross-reference after those dates.

However, I do not understand why it still fails (even in 5.5) with the first two variants.

Kristian Nielsen added a comment - 2015-01-06 12:54 I think the third one is already fixed, in revid:knielsen@knielsen-hq.org-20141031114817-art3oxugmfgfd6ca This fix appeared in 5.5 around 2014-10-31 and in 10.0 around 2014-11-03, and there are no failures of that kind seen in cross-reference after those dates. However, I do not understand why it still fails (even in 5.5) with the first two variants.

Kristian Nielsen added a comment - 2015-01-06 17:05

Ok, I found it.

The test fails only in --ps-protocol (variation 2). Cross-reference shows some
failures without --ps-protocol (variation 1), but these are in trees where the
existing fix in 5.5 was not yet merged at the time of the failure. Same for
variation 3, also only in trees in which the fix is not yet merged from 5.5.

The problem is that the test case sets up a DEBUG_SYNC that is meant to
trigger at the end of the execution of the SELECT SLEEP(5). But in
--ps-protocol, a query happens in two steps, Prepare followed by Execute. And
the DEBUG_SYNC ends up triggering at the end of the Prepare step, which is too
early, allowing a race to cause the test to fail sporadically.

Kristian Nielsen added a comment - 2015-01-06 17:05 Ok, I found it. The test fails only in --ps-protocol (variation 2). Cross-reference shows some failures without --ps-protocol (variation 1), but these are in trees where the existing fix in 5.5 was not yet merged at the time of the failure. Same for variation 3, also only in trees in which the fix is not yet merged from 5.5. The problem is that the test case sets up a DEBUG_SYNC that is meant to trigger at the end of the execution of the SELECT SLEEP(5). But in --ps-protocol, a query happens in two steps, Prepare followed by Execute. And the DEBUG_SYNC ends up triggering at the end of the Prepare step, which is too early, allowing a race to cause the test to fail sporadically.

Kristian Nielsen added a comment - 2015-01-06 17:36

Pushed to 5.5.42:

http://lists.askmonty.org/pipermail/commits/2015-January/007230.html

Note that the test failures might be visible in Buildbot for some time yet,
until the fix becomes merged from 5.5 into each tree.

Kristian Nielsen added a comment - 2015-01-06 17:36 Pushed to 5.5.42: http://lists.askmonty.org/pipermail/commits/2015-January/007230.html Note that the test failures might be visible in Buildbot for some time yet, until the fix becomes merged from 5.5 into each tree.

Kristian Nielsen made changes - 2015-01-06 17:36

Fix Version/s		N/A [ 14700 ]
Fix Version/s	5.5 [ 15800 ]
Fix Version/s	10.0 [ 16000 ]
Fix Version/s	10.1 [ 16100 ]
Resolution		Fixed [ 1 ]
Status	In Progress [ 3 ]	Closed [ 6 ]

Elena Stepanova made changes - 2015-01-08 01:46

Fix Version/s		10.1.3 [ 18000 ]
Fix Version/s		10.0.16 [ 17900 ]
Fix Version/s		5.5.42 [ 18102 ]
Fix Version/s	N/A [ 14700 ]

Rasmus Johansson (Inactive) made changes - 2015-05-18 17:51

Workflow

MariaDB v2 [ 58707 ]

MariaDB v3 [ 65491 ]

Sergei Golubchik made changes - 2021-12-06 21:40

Workflow

MariaDB v3 [ 65491 ]

MariaDB v4 [ 148536 ]

People

Assignee:: Kristian Nielsen

Reporter:: Elena Stepanova

Votes:: 0 Vote for this issue

Watchers:: 2 Start watching this issue

Dates

Created:: 2014-11-24 16:08

Updated:: 2015-01-08 01:46

Resolved:: 2015-01-06 17:36

Git Integration

Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.

MariaDB Server

Details

Description

Attachments

Issue Links

Activity

People

Dates

Git Integration