On Fri, Nov 24, 2006 at 04:49:00PM +0500, Igor Goldenberg wrote:
> Hello,
>
> I have a question about RAIDframe driver (using in RAID1 mode).
>
> GENERIC kernel doesn't support it, so it's need to compile custom
> kernel. As described in FAQ it's not recommended. Also additional
> complexity for kernel is not good as I think. And upgrade from one
> release to another (the next) will be more difficult.
>
> Also if server was incorrectly shutdown parity syncronization requires
> extra time while boot.
>
> Is RAIDframe driver stable enough for production server (for custom
> partition like /var/flow and system partitions as for /usr, /var
> except /)? Or it's better just to use backup of root partition (using
> cron + /altroot mechanism) and something like rsync to another local
> harddisk for netflow logs?
>
> If I use small amount of binary packets (like netflow-tools, pfflowd,
> postgresql client) and backup root partition and custom data as
> mention above will be reinstall system from scratch simple?
>
> So, what's better - to have base system partially on RAID or only for
> custom data or not to use RAID at all?
This is just one opinion, but...
I do not RAID the system. All binaries are easily reconstructed from
mirrors/ports, and /etc doesn't change often enough that restoring the
daily backup is likely to be a problem.
On some, but not all, boxes I do use /altroot and similar for usr and
home (this is a server, so /home doesn't get a lot of action).
I *do* use RAIDframe to protect rapidly-changing data - for instance,
/var/mail on mailservers.
Firewalls are greatly simplified by the introduction of a loghost; I
have one, and don't bother RAID'ing the firewall. If I do need more
reliability, I'll use CARP.
Joachim
