FreeBSD 6.0-release AMD64 stability/reliability?
kgunders at teamcool.net
Tue Nov 22 18:20:56 GMT 2005
On Tue, 22 Nov 2005 10:06:53 +0100
Olaf Greve <o.greve at axis.nl> wrote:
> Well, the people who are going to order the hardware for the DB server I
> have been asking some question about have almost made their minds up. It
> seems indeed they'll go for an SATA 3ware RAID controller... Depending
> on availability and price locally over here they'll probably choose
> something from the 8000 or 9000 series. We'll see...
> Meanwhile: indeed I will be the one who will be installing the OS, and
> I'm looking at FreeBSD 6.0-release AMD64.
> Can anyone tell me whether that already is a smart choice for a live DB
> server (in terms of matureness, stability, reliability, etc.), or
> whether it would be wiser to stick to FreeBSD 5.4-release AMD64 for this?
> Also: in case at a later point in time I want to upgrade the complete OS
> to a more recent version (once they're available), will it be easy to go
> through the 'buildworld' process? From what I gather this shouldn't be
> too much of an issue, it's just that I've never done it before...
> Tnx in advance for any answers, and cheers!
fwiw-- I've been using FreeBSD since 2.x days. During that time I've
run into a few minor glitches upgrading. But that is so infrequent that
I couldn't even really recount them. Usually a fresh cvsup and a couple
'make cleandir' did the trick. Of course you had to reinstall from
scratch to upgrade from UFS1 based system to UFS2, but other than that
I can't recall a forced reinstall on a production box. Maintainability
and upgradability are one the the hallmark features that make FreeBSD
>From my recollection there were a few issues w/early 3.x releases. 4.x
were pretty solid from the get go. Of course 5.x was very ambitious
and reaching stable status took a bit of doing, during which time 4.x
was actively maintained and many 5.x improvements back ported.
The path to 5.x stable was atypical and I think the FBSD developers
took the opportunity to learn from the experience. FreeBSD's kernel and
userland are an integrated and cohesive package rather than separate
entities. Overall upgrades are pretty smooth. FreeBSD may run into
a glitch or two here and there between major releases but nothing like
the horror stories you may have heard about with some other *nix like
operating sytems. The point being here that I think FreeBSD has done a
damn good job of living up to it's motto. First and foremost FreeBSD
provides and excellent server platform.
Presently I've got my production machines on 5.4. I'm doing new builds
on 6.0. They haven't been actually deployed into production yet due to
time constraints of 3rd parties. I've not done any formal benchmarking
but my informal assessment is that they "feel" a bit "snappier" than 5.4
on comparable hardware. Release engineering did an excellent job w/the
6.0 release. I suspect Scott's testing proceedures contributed
significantly in this regard. Hence, I'm feeling pretty optimistic
about 6.0 in production.
I've only upgraded one test machine from 5.4 to 6.0. Don't recall
encountering any growing pains. Be sure to check /usr/src/UPDATING.
I suggest you get yourself a low cost test box that you can use for
testing. Clone your production machine config to the extent you
are able, but any old workstation will do in a pinch. Then before you
upgrade/update anything on the production machine, test on the
practice machine first. You may run into undocumented incompatabilites
during ports upgrades but once again, these are pretty infrequent. Be
sure to keep an eye on /usr/ports/ UPDATING as well.
Q: Because it reverses the logical flow of conversation.
A: Why is putting a reply at the top of the message frowned upon?
More information about the freebsd-amd64