<div dir="ltr"><br><div>We are seeing some of our apache2 instances with shibboleth leaking log file descriptors. apache eventually runs out of fd's and temporarily rejects incoming connections. It looks like there are thousands of native.log fd's that are deleted but still owned by apache2. </div><div><br></div><div><span style="font-family:Monaco;font-size:11px"># lsof | grep native.log | grep deleted </span><br></div><div>(snippet)</div><div><p style="margin:0px;font-size:11px;font-family:Monaco">apache2 24990 httpd 418w REG 8,3 1001778 749740 /usr/local/httpd/logs/shibboleth-sp/native.log.6 (deleted)</p><p style="margin:0px;font-size:11px;font-family:Monaco">apache2 24990 httpd 420w REG 8,3 176 749856 /usr/local/httpd/logs/shibboleth-sp/native.log.10 (deleted)</p><p style="margin:0px;font-size:11px;font-family:Monaco">apache2 24990 httpd 422w REG 8,3 1001778 749762 /usr/local/httpd/logs/shibboleth-sp/native.log.10 (deleted)</p><p style="margin:0px;font-size:11px;font-family:Monaco"><br></p></div><div><p style="margin:0px;font-size:11px;font-family:Monaco"># lsof | grep native.log | grep deleted | wc -l</p>
<p style="margin:0px;font-size:11px;font-family:Monaco">3612</p></div><div><br></div><div>Some cursory debugging shows that apache2 is creating multiple instances of the RollingFileAppender object which I do not believe the log4* protects correctly during a roll over.</div><div>Has anyone experienced such problems using the default native.logging config?</div><div><br></div><div> Thanks</div></div>