[nSLUG] (2) Bloat & Notice of Upgrades
nslug at fop.ns.ca
Thu Sep 9 14:13:15 ADT 2004
On Thu, 9 Sep 2004, David Potter wrote:
> The business expects to be able to operate 7x16 and during these work
> hours the fileserver is what I consider to be mission critical - if it
> fails, everything stops <period>.
My experience, FWIW, is RedHat is somewhat unreliable and flaky. I've had
bad experiences with FC2 (just... broken in *so* many ways), RH9
(gcc/library brokenness, rsync failing for no particularly obvious reason,
mailman broken), RH8 (security holes), and RH7.* and prior (excessively
clunky). On top of that there's the whole RPM dependency hell, which is
somewhat ameliorated by the porting of apt-get and friends from Debian.
That said, if you want to run a mission critical server, then the issue
isn't so much the underlying OS as preparedness for disasters; I would
personally use mirrored drives on the fileserver, and also run it as a
cluster using something like DRBD to give RAID1-over-network capabilities.
That should give you reasonably high availability without having to go for
something extremely expensive like a SAN.
> Having said that, I manage servers rather conservatively, and I've never
> had downtime that was caused by either the OS or the application
> software - which is why the servers are Linux and not MS.
Downtime on the servers I've managed over the past year has been something
1 hour: Waiting for Windows servers to come back up after patching.
35 minutes: Outage due to Juan, UPS batteries weren't quite able to cover
the outage till power came back.
120 seconds: Testing cluster failover once a month.
Unfortunately, Win2K is required for some of our clients...
More information about the nSLUG