Bob Friesenhahn wrote: > On Thu, 5 Jun 2008, Richard Elling wrote: > > >> Nathan Kroenert wrote: >> >>> I'd expect it's the old standard. >>> >>> if /var/tmp is filled, and that's part of /, then bad things happen. >>> >> Such as? If you find a part of Solaris that cannot deal with a full >> file system, then please (pretty please... with a cherry on top) file a >> bug. >> > > I expect that it matters when the filesystem gets filled up. The > functionality offered by /var/run would become quite broken if it was > full during boot. No nameservice. No daemon PID files. Doors > broken. Boo-Hoo. >
/var/run is mounted in virtual memory and not disk. So it doesn't matter if the /var filesystem is full or not. OTOH, if you are out of virtual memory, then I agree there is lots of stuff that would fail. So, we had this argument about 7-8 years ago and to prove it we setup a server providing many services (http, NFS) and wrote a script which would keep / full (not as trivial as you might think). We ran for 6 months before we decided to cancel the experiment (point made, and we got bored :-). We have not repeated the experiment with Solaris 10 or NV, though, so I can't say that there aren't any bugs that might have sneaked in. If you find one, then please file a bug. Note: off-alias, someone pointed out a potential bug, so please keep those cards and letters coming! We want Solaris to be much more robust! -- richard _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
