Claus Guttesen a ?crit :> Hi.
Hi,
>
> Last week I phased out my remaining trusty FreeBSD nfs-server. The
> first was phased out three weeks ago. They have served me well for
> about two year and I have been *very* satisfied with the performance
> and stability. The below mentioned reasons made the decision easier to
> migrate to veritas volume manager on solaris.
>
> 1. Lack of a journaling filesystem.
> 2. Lack of a logical volume manager.
>
> My intent is *not* to start a flamewar, simpy stating why I had to migrate.
>
> Additional comments to item #1:
> We have background-fsck, what's wrong with that?
>
> Well, there is nothing wrong with that in a way. Background-fsck does
> work, but my nfs-server have had three unplanned reboots during that
> time, none of them was caused by the nfs-server itself, but caused by
> other factors. The server comes back up as it should and detects that
> the volumes wasn't unmounted in an orderly fashion and defers the
> volume to background-fsck. So far so good.
>
> When the background-fsck is done with one volume and it jumps to the
> next, my webservers connected to the nfs-server are unable to read and
> write to the nfs-volumes for a period of 15-30 minutes. The smallest
> (of several) volume is 400 GB and the largest (of several) is 2 TB.
> The outcome is that my website is seen as being inaccessible. This was
> with FreeBSD 5.2 beta through 5.4 I saw this behaviour.
we switched our mail server to Linux for your first reason :
see
http://lists.freebsd.org/mailman/htdig/freebsd-stable/2005-June/thread.html#15751
>
> So I'm delighted to read that the initial work on a journaling
> filesystem has started.
>
> Additional comments to item #2:
> Use vinum!
>
> Is it vinum or gvinum which is the future of FreeBSD?
>
> The docs related to vinum refers to some parameters in newfs not
> present in the manual-pages.
>
> As more volumes are added the task of configuring (g)vinum will become
> more and more timeconsuming and errorprone. Does it recover correctly
> in the event of a crash, how about fsck/newfs on volumes larger than 2
> TB?
>
>
> The camcontrol program on FreeBSD is a very robust tool. This is one
> program I miss. Some parameters to the find- and date-commands on
> FreeBSD aren't there on solaris, so I'll keep the old nfs-server
> around for doing day2day maintenance.
>
> I'm keeping FreeBSD as webservers (of course).
>
> regards
> Claus
--
Philippe PEGON