I posted this in the General Problems forum and am re-posting here.
Our environment has been having sporadic issues where pages are taking ten seconds to several minutes to load at random intervals throughout the day. If usage is particularly heavy, requests pile up, the DB max connections pool is exceeded, and users begin receiving error messages.
I have run strace on httpd processes during these outages and I am attaching the output of five processes that took between 30 seconds and three minutes to complete a page request. My analysis of these indicates that the "log jam" is occurring when Moodle is trying to access files in the moodledata folder, which is stored on a separate server and mounted via NFS. In each of these 5 cases, the file/folder being accessed is /mnt/moodledata/temp/typo3temp/cs, which is being requested by /lib/typo3/class.t3lib_div.php.
A few questions- first of all is that hypothesis correct? I am basing this on the timestamps in the traces before and after the request for /mnt/moodledata/temp/typo3temp/cs is made. Second, could this particular file/library (typo3) be the cause of this problem, or is it more likely to be a server/network/NFS issue? If it is not a Moodle issue, what are some linux tools available that could help pinpoint the problem further, or other steps that may help mitigate or better manage this problem? Restarting apache does resolve it temporarily, but that is not a viable long term solution.