[JIRA] Commented: (SSPCPP-523) Strange behaviour of native log

[Scott Cantor (JIRA)]

    [ https://issues.shibboleth.net/jira/browse/SSPCPP-523?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14568#comment-14568 ] 

Scott Cantor commented on SSPCPP-523:
-------------------------------------

I know prefork is the default, what I'm asking is, does using worker fix this?

On every copy I have, the file descriptors are inherited. I wonder though if it's caused by the rollover that log4cpp is doing, which Apache doesn't know about. Maybe it looks like it's working for me initially but then fails later.

The problem with that theory is that I think they are seeing this break right away.

The short answer is that if root ownership isn't working, then nothing has really changed, and just like before, they have to fix the permissions. I can't do anything about that short of not logging anything.

But I want to understand why it appears to be working for me.

> Strange behaviour of native log 
> --------------------------------
>
>                 Key: SSPCPP-523
>                 URL: https://issues.shibboleth.net/jira/browse/SSPCPP-523
>             Project: Shibboleth SP - C++
>          Issue Type: Bug
>            Reporter: Lukas Hämmerle
>            Assignee: Scott Cantor
>            Priority: Minor
>
> One of our universities observed some strange behaviour of the native.log. With RedHat 5 they only hardly get any entries besides the ones below when starting the SP:
> 	2012-11-04 20:27:12 INFO XMLTooling.Config : xmltooling 1.5.1 library initialization complete
> 	2012-11-04 20:27:12 INFO Shibboleth.Config : shibboleth 2.5.0 library initialization complete
> 	2012-11-04 20:27:12 INFO Shibboleth.Config : shibboleth 2.5.0 library shutting down
> 	2012-11-04 20:27:12 INFO XMLTooling.Config : xmltooling 1.5.1 library shutdown complete
> 	2012-11-04 20:27:12 INFO Shibboleth.Config : shibboleth 2.5.0 library shutdown complete
> 	2012-11-04 20:27:12 INFO XMLTooling.Config : xmltooling 1.5.1 library initialization complete
> 	2012-11-04 20:27:12 INFO Shibboleth.Config : shibboleth 2.5.0 library initialization complete
> 	2012-11-04 20:27:12 INFO Shibboleth.Config : reload thread started...running when signaled
> 	2012-11-04 20:27:12 INFO Shibboleth.Config : reload thread started...running when signaled
> However, there should be more entries than this (in my centos 5 installation about 191 INFO entries)
> So, they added <InProcess logger="native.logger"/> to shibboleth2.xml, which by default is not included.
> This then results in the following CRIT entries:
> 	2012-11-04 20:55:04 CRIT XMLTooling.Logging : error in file permissions or logging configuration: exception creating appender: failed to open log file (/var/log/httpd/native.log)
> 	2012-11-04 20:55:04 CRIT Shibboleth.Config : failed to load new logging configuration from (native.logger)
> However, the file /var/log/httpd/native.log exists with root:root 644. It also does not play a difference if the native.log is written to /tmp/. Same result. Also, in this case the log rotation does not work and the log file grows larger and larger. After a while one will find exclusively entries like these:
> 	2012-11-07 13:46:17 INFO Shibboleth.Config : loaded XML resource (/etc/shibboleth/shibboleth2.xml)
> 	2012-11-07 13:46:45 INFO Shibboleth.Config : reload thread started...running when signaled
> Something else they observed are many entries like:
> 2012-11-06 04:25:46 WARN Shibboleth.ServiceProvider [27347] shib_check_user: can't export assertions without an exportLocation Sessions property
> Is it justified to make these WARN entries if exportAssertion is not used at all?

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira