There is a lot of email in the archives of this list complaining about things such as warning: unable to unlink local/9/3601004; will try again later I saw this too, (running with the rpms made by Bruce Guenter E<[EMAIL PROTECTED]>) I investigated what was going on. The key is to look at errno when the unlink fails. (By the way, I suggest that the when printing the warning about the unlink failing, the error code ought to be printed out too.) The unlink returned error code 5 (I/O error) sometimes, but not always. By taking out the syncdir patch, the problem goes away. I mounted the ext2fs filesystem "sync" (it turns out the only thing on that disk is my qmail queue and my alias maildirs, so it is an excellent candidate for being mounted "sync".) Now the system works much better, with none of those "unable to unlink" messages in the logs. A related problem: The "try again later" is 123 seconds later: pe.dt = now() + SLEEP_SYSFAIL; This can cause problem if more than a few hundred messages get into this state (especially when using syslogd). The problem is that qmail spends all its time looking at these messages. Much better would be if the retry were scheduled with a quadratic backoff strategy to avoid swamping qmail with these bad messages. -Bradley
