You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@trafficserver.apache.org by "Leif Hedstrom (JIRA)" <ji...@apache.org> on 2011/09/15 17:44:09 UTC
[jira] [Created] (TS-955) TS-168 breaks regressions for TextLog
TS-168 breaks regressions for TextLog
-------------------------------------
Key: TS-955
URL: https://issues.apache.org/jira/browse/TS-955
Project: Traffic Server
Issue Type: Bug
Components: Logging
Affects Versions: 3.1.1
Reporter: Leif Hedstrom
Fix For: 3.1.1
With the fixes from TS-168, the logging regressions can fail if you run the traffic_server -R 1 more than once. The first run always succeeds, but the 2nd and subsequent run can fail. What seems to happen is that the (small) log is not flushed, and the log is not created until the first flush happens. So, everything looks like it works, up until we (5s after log creation) try to read the log. The log then doesn't exist, and the regression in log_test_handler() fails (since, the file can't be open nor read).
I've tracked this down to the commit for TS-168, and also traced through the test in gdb, and as far as I can tell, the flush never happens, which means the log write never happens, and hence, the log file creation never happens either.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Work started] (TS-955) TS-168 breaks regressions for
TextLog
Posted by "Zhao Yongming (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-955?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Work on TS-955 started by Zhao Yongming.
> TS-168 breaks regressions for TextLog
> -------------------------------------
>
> Key: TS-955
> URL: https://issues.apache.org/jira/browse/TS-955
> Project: Traffic Server
> Issue Type: Bug
> Components: Logging
> Affects Versions: 3.1.1
> Reporter: Leif Hedstrom
> Assignee: Zhao Yongming
> Priority: Critical
> Fix For: 3.1.1
>
> Attachments: TS-955.patch
>
>
> With the fixes from TS-168, the logging regressions can fail if you run the traffic_server -R 1 more than once. The first run always succeeds, but the 2nd and subsequent run can fail. What seems to happen is that the (small) log is not flushed, and the log is not created until the first flush happens. So, everything looks like it works, up until we (5s after log creation) try to read the log. The log then doesn't exist, and the regression in log_test_handler() fails (since, the file can't be open nor read).
> I've tracked this down to the commit for TS-168, and also traced through the test in gdb, and as far as I can tell, the flush never happens, which means the log write never happens, and hence, the log file creation never happens either.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (TS-955) TS-168 breaks regressions for TextLog
Posted by "Zhao Yongming (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-955?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zhao Yongming updated TS-955:
-----------------------------
Attachment: TS-955.patch
there is something need to know:
the flashing is in another LOGGING thread, and that is async from the testing
in the testing codes, the log file checking is done after the log object is destroyed, which follows right after the flushing. that will make some very candy situation where when the destroy may cancel the flushing.
here is the patch which will fix:
1, make destroy after the log file checking, increase the time wait for flushing from 5s to 6s.
2, cleanup the log files before and after the testing, we should not mess up the logging dir.
it is dirty.
> TS-168 breaks regressions for TextLog
> -------------------------------------
>
> Key: TS-955
> URL: https://issues.apache.org/jira/browse/TS-955
> Project: Traffic Server
> Issue Type: Bug
> Components: Logging
> Affects Versions: 3.1.1
> Reporter: Leif Hedstrom
> Assignee: Zhao Yongming
> Priority: Critical
> Fix For: 3.1.1
>
> Attachments: TS-955.patch
>
>
> With the fixes from TS-168, the logging regressions can fail if you run the traffic_server -R 1 more than once. The first run always succeeds, but the 2nd and subsequent run can fail. What seems to happen is that the (small) log is not flushed, and the log is not created until the first flush happens. So, everything looks like it works, up until we (5s after log creation) try to read the log. The log then doesn't exist, and the regression in log_test_handler() fails (since, the file can't be open nor read).
> I've tracked this down to the commit for TS-168, and also traced through the test in gdb, and as far as I can tell, the flush never happens, which means the log write never happens, and hence, the log file creation never happens either.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-955) TS-168 breaks regressions for TextLog
Posted by "Leif Hedstrom (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-955?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13106836#comment-13106836 ]
Leif Hedstrom commented on TS-955:
----------------------------------
Removing the old log files (from the previous run) "fixes" the problem, so perhaps something around log file rotations or something ?
> TS-168 breaks regressions for TextLog
> -------------------------------------
>
> Key: TS-955
> URL: https://issues.apache.org/jira/browse/TS-955
> Project: Traffic Server
> Issue Type: Bug
> Components: Logging
> Affects Versions: 3.1.1
> Reporter: Leif Hedstrom
> Assignee: Zhao Yongming
> Priority: Critical
> Fix For: 3.1.1
>
>
> With the fixes from TS-168, the logging regressions can fail if you run the traffic_server -R 1 more than once. The first run always succeeds, but the 2nd and subsequent run can fail. What seems to happen is that the (small) log is not flushed, and the log is not created until the first flush happens. So, everything looks like it works, up until we (5s after log creation) try to read the log. The log then doesn't exist, and the regression in log_test_handler() fails (since, the file can't be open nor read).
> I've tracked this down to the commit for TS-168, and also traced through the test in gdb, and as far as I can tell, the flush never happens, which means the log write never happens, and hence, the log file creation never happens either.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-955) TS-168 breaks regressions for TextLog
Posted by "Zhao Yongming (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-955?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13105981#comment-13105981 ]
Zhao Yongming commented on TS-955:
----------------------------------
I get a log test fail with:
{code}
[SDK_API_TSTextLog] TSTextLogObject : [TestCase1] <<FAIL>> { can not open log file var/log/trafficserver/RegressionTestLog25248.log }
{code}
> TS-168 breaks regressions for TextLog
> -------------------------------------
>
> Key: TS-955
> URL: https://issues.apache.org/jira/browse/TS-955
> Project: Traffic Server
> Issue Type: Bug
> Components: Logging
> Affects Versions: 3.1.1
> Reporter: Leif Hedstrom
> Assignee: Zhao Yongming
> Priority: Critical
> Fix For: 3.1.1
>
>
> With the fixes from TS-168, the logging regressions can fail if you run the traffic_server -R 1 more than once. The first run always succeeds, but the 2nd and subsequent run can fail. What seems to happen is that the (small) log is not flushed, and the log is not created until the first flush happens. So, everything looks like it works, up until we (5s after log creation) try to read the log. The log then doesn't exist, and the regression in log_test_handler() fails (since, the file can't be open nor read).
> I've tracked this down to the commit for TS-168, and also traced through the test in gdb, and as far as I can tell, the flush never happens, which means the log write never happens, and hence, the log file creation never happens either.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-955) TS-168 breaks regressions for TextLog
Posted by "Zhao Yongming (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-955?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13114268#comment-13114268 ]
Zhao Yongming commented on TS-955:
----------------------------------
r1175400
> TS-168 breaks regressions for TextLog
> -------------------------------------
>
> Key: TS-955
> URL: https://issues.apache.org/jira/browse/TS-955
> Project: Traffic Server
> Issue Type: Bug
> Components: Logging
> Affects Versions: 3.1.1
> Reporter: Leif Hedstrom
> Assignee: Zhao Yongming
> Priority: Critical
> Fix For: 3.1.1
>
> Attachments: TS-955.patch
>
>
> With the fixes from TS-168, the logging regressions can fail if you run the traffic_server -R 1 more than once. The first run always succeeds, but the 2nd and subsequent run can fail. What seems to happen is that the (small) log is not flushed, and the log is not created until the first flush happens. So, everything looks like it works, up until we (5s after log creation) try to read the log. The log then doesn't exist, and the regression in log_test_handler() fails (since, the file can't be open nor read).
> I've tracked this down to the commit for TS-168, and also traced through the test in gdb, and as far as I can tell, the flush never happens, which means the log write never happens, and hence, the log file creation never happens either.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Resolved] (TS-955) TS-168 breaks regressions for TextLog
Posted by "Zhao Yongming (Resolved) (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-955?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zhao Yongming resolved TS-955.
------------------------------
Resolution: Fixed
> TS-168 breaks regressions for TextLog
> -------------------------------------
>
> Key: TS-955
> URL: https://issues.apache.org/jira/browse/TS-955
> Project: Traffic Server
> Issue Type: Bug
> Components: Logging
> Affects Versions: 3.1.1
> Reporter: Leif Hedstrom
> Assignee: Zhao Yongming
> Priority: Critical
> Fix For: 3.1.1
>
> Attachments: TS-955.patch
>
>
> With the fixes from TS-168, the logging regressions can fail if you run the traffic_server -R 1 more than once. The first run always succeeds, but the 2nd and subsequent run can fail. What seems to happen is that the (small) log is not flushed, and the log is not created until the first flush happens. So, everything looks like it works, up until we (5s after log creation) try to read the log. The log then doesn't exist, and the regression in log_test_handler() fails (since, the file can't be open nor read).
> I've tracked this down to the commit for TS-168, and also traced through the test in gdb, and as far as I can tell, the flush never happens, which means the log write never happens, and hence, the log file creation never happens either.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-955) TS-168 breaks regressions for TextLog
Posted by "Leif Hedstrom (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-955?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13106067#comment-13106067 ]
Leif Hedstrom commented on TS-955:
----------------------------------
Right. That happens because the write on the log doesn't seem to get flushed. And therefore the text log file is never created, so the test to read the log fails (since the filevdoesnt exist).
> TS-168 breaks regressions for TextLog
> -------------------------------------
>
> Key: TS-955
> URL: https://issues.apache.org/jira/browse/TS-955
> Project: Traffic Server
> Issue Type: Bug
> Components: Logging
> Affects Versions: 3.1.1
> Reporter: Leif Hedstrom
> Assignee: Zhao Yongming
> Priority: Critical
> Fix For: 3.1.1
>
>
> With the fixes from TS-168, the logging regressions can fail if you run the traffic_server -R 1 more than once. The first run always succeeds, but the 2nd and subsequent run can fail. What seems to happen is that the (small) log is not flushed, and the log is not created until the first flush happens. So, everything looks like it works, up until we (5s after log creation) try to read the log. The log then doesn't exist, and the regression in log_test_handler() fails (since, the file can't be open nor read).
> I've tracked this down to the commit for TS-168, and also traced through the test in gdb, and as far as I can tell, the flush never happens, which means the log write never happens, and hence, the log file creation never happens either.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-955) TS-168 breaks regressions for TextLog
Posted by "Leif Hedstrom (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-955?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13106066#comment-13106066 ]
Leif Hedstrom commented on TS-955:
----------------------------------
Right, that happens because the write to that log is never flushed, so the log file is never created.
> TS-168 breaks regressions for TextLog
> -------------------------------------
>
> Key: TS-955
> URL: https://issues.apache.org/jira/browse/TS-955
> Project: Traffic Server
> Issue Type: Bug
> Components: Logging
> Affects Versions: 3.1.1
> Reporter: Leif Hedstrom
> Assignee: Zhao Yongming
> Priority: Critical
> Fix For: 3.1.1
>
>
> With the fixes from TS-168, the logging regressions can fail if you run the traffic_server -R 1 more than once. The first run always succeeds, but the 2nd and subsequent run can fail. What seems to happen is that the (small) log is not flushed, and the log is not created until the first flush happens. So, everything looks like it works, up until we (5s after log creation) try to read the log. The log then doesn't exist, and the regression in log_test_handler() fails (since, the file can't be open nor read).
> I've tracked this down to the commit for TS-168, and also traced through the test in gdb, and as far as I can tell, the flush never happens, which means the log write never happens, and hence, the log file creation never happens either.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (TS-955) TS-168 breaks regressions for TextLog
Posted by "Leif Hedstrom (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/TS-955?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Leif Hedstrom updated TS-955:
-----------------------------
Priority: Critical (was: Major)
Assignee: Zhao Yongming
> TS-168 breaks regressions for TextLog
> -------------------------------------
>
> Key: TS-955
> URL: https://issues.apache.org/jira/browse/TS-955
> Project: Traffic Server
> Issue Type: Bug
> Components: Logging
> Affects Versions: 3.1.1
> Reporter: Leif Hedstrom
> Assignee: Zhao Yongming
> Priority: Critical
> Fix For: 3.1.1
>
>
> With the fixes from TS-168, the logging regressions can fail if you run the traffic_server -R 1 more than once. The first run always succeeds, but the 2nd and subsequent run can fail. What seems to happen is that the (small) log is not flushed, and the log is not created until the first flush happens. So, everything looks like it works, up until we (5s after log creation) try to read the log. The log then doesn't exist, and the regression in log_test_handler() fails (since, the file can't be open nor read).
> I've tracked this down to the commit for TS-168, and also traced through the test in gdb, and as far as I can tell, the flush never happens, which means the log write never happens, and hence, the log file creation never happens either.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira