pion: I can't prove anything from those worker exit debug lines, but assuming that those are all the lines it's REALLY interesting that the dates you were having multiple lockups are exactly the dates you were seeing the highest load. That might mean that thread creation, which requires the dll lock, is equally likely to trigger the lockup. Try upping the number of worker threads to like 10 and see if that makes a difference!
Rusher: That means that your CPU must be VERY busy or insane numbers of control connections all at one time. Try looking for other lines in error.log or debug.log that look like:
Services_Test: Failed to connect to service '%s' (IP=%s) %d times in a row!
ioFTPD clearly thought it couldn't connect to the FTP service defined in the .ini file (not a windows services). Since there was only 1 it figured it should suicide...
That can happen under high CPU load or network load on the control connection. If this is happening I can tweak the timeouts to avoid this for other people or to disable the feature. The ONLY times I've personally seen this is when older machines burn CD/DVDs or play high def video. Any chance of that kind of thing going on?
There is a workaround in the short term, you define a 2nd local-only FTP service. Now if one of them appears offline the server notes it unless under extreme stress the local only one shouldn't fail...
In the future I'll adjust thread priorities to avoid this perhaps.
|