shib native logging with RollingFileAppender

Cantor, Scott cantor.2 at osu.edu
Wed Sep 16 10:46:38 EDT 2015


On 9/16/15, 2:08 AM, "users on behalf of Jeremy Shapiro" <users-bounces at shibboleth.net on behalf of jnshapiro at gmail.com> wrote:



>OK, so that's not entirely true.  We're not making shib config changes but we are HUP'ing apache2 for httpd.conf changes every few days.

I was mainly referring to the SP's config, but mostly out of concern that that actually was causing a problem it shouldn't be. Using the logger attributes causes the logging configuration to be explicitly redone every time the SP configuration is reloaded.

>I see that this causes the shib module to get reconfigured.  If this reinitialization and the associated logs generated are enough to cause native.log to rollOver, then we get the leaked deleted fd's in the lsof output.

I would imagine, since my server's not showing that result and doesn't get reloaded like that. A full reload of the module looks like a from-scratch startup to the libraries underneath.

The logging library has no facility for "shutting down", there's no way to get it to destroy its objects other than ending the process. Either turn off the log, or try using the syslog appender.

-- Scott



More information about the users mailing list