On 2006-10-30 at 20:59 -0500, Grant Peel wrote: > every once in a while, I get a mail account that I have to delete and reinstall ... here is the indicator:
>
> 2006-10-30 05:49:52 1GeSHv-000A5B-Do == AAA@??? R=virtual_localuser T=virtual_localdelivery defer (-9): failed to lock mailbox /var/spool/virtual/BBBB.com/AAAA (fcntl/flock)
>
> Oddly enough, when this happens, there is no .lock file to be found.
There won't be a .lock file when fcntl/flock are used. Those are
in-kernel metadata associated with the file.
One option is lsof/fuser/whichever with the filename to see what has it
open.
Another option is that you have a kernel bug and locks have been leaked.
Hopefully this is much less likely than that another process has the
lock open. Locks _shouldn't_ be able to linger past process exit
because they're automatically released when the open file-descriptor is
closed and the kernel ensures that this happens, no matter how the
process ends.
> Again, all that seems to work is to delete the account and reinstall it.
So that removes the file, including any locks help on it. So the
lock-count on the file was non-zero. So lsof(8) or fuser(1/1m) or
whichever your OS provides is the first step to identifying the rogue
process.
