ralfthewise wrote:
> Done a lot of research on this one and can't seem to find the solution.
> I have a linux server, runs apache and serves a bunch of different
> virtual servers, some of which have their data stored on other windows
> machines, which I mount from the linux server via cifs. If I shutdown
> one of the windows machines that is hosting the files for one of the
> virtual servers, obviously that mount point is dead on the linux
> server.
umount first
> This doesn't really matter at first, it just means that
> virtual server stops working. If, however, I now try to stop and start
> apache on the linux server, well a few httpd processes don't stop
> because they are trying to use the dead mount point, and then apache
> won't start because there are already processes running that are
> listening on port 80. The only way I can unmount the dead mount points
> is to do a 'umount -l', but the zombie httpd processes still keep
> running after doing the lazy unmount.
Did the `umount -l' really unmounted at that point?
Are you sure these procs are zombies (defunct state) but not rather
in uninterruptable sleep state?
Zombies are dead processes appearing only in the process table, ie.
consuming no system resources.
--
reply to usenet only
|