[nycbug-talk] [OT] FreeBSD 5.4 crashing?

[pete wright]

On 6/30/05, Matt Juszczak <matt at atopia.net> wrote:
> Hi all,
> 
> The past two weeks have been rough for me.  I've had our production mail
> server, FreeBSD 5.4, and our back end LDAP servers crashing.  I've tried
> everything, from upgrading to 5.4-STABLE, and now switching from IPF to PF
> (which actually has made them not crash).
> 
> Problem is, even if these boxes now stay up for a month, I don't know if I
> can trust my FreeBSD 5.4 boxes and sleep well at night (at least not
> these).  I think in order to satisfy my boss as well (he wants to switch
> to Debian which I'm struggling to aim away from for a high traffic mail
> server), I need to switch OS's to something that I know will be stable so
> we can get our new LDAP implementation in place and I can move on to other
> projects.....
> 
> My debate right now is FreeBSD 4.11 or OpenBSD 3.7.  As an avid and long
> time FreeBSD user, my recent install of OpenBSD seemed to go smoothly.
> Problem is, we use adaptec raid, and OpenBSD recently lost support for
> that, but apparently a simple kernel rebuild can do the trick.  Also,
> people have told me that, although the SMP code is more stable on OpenBSD
> (some responses have told me that the reason my FreeBSD installs seem to
> keep kernel trapping HARD (they lock up) is because of unstable SMP code
> in 5.x), it isn't used as efficiently.
> 
> Therefore .... what do you all think?  Our setup, to recap, is like this:
> 
> Orion ---> Mail server, dual 3.06 ghz xeon, 4 GB DDR 2100 Kingston, LDAP.
> Uranus --> LDAP server, dual 850 MHZ PIII, 2 GB DDR, regular PAM.
> Caliban -> LDAP server, dual 850 MHZ PIII, 2 GB DDR, regular PAM.
> 

couple issues I would ask before deciding to move to another OS all
together.  what where your previous production systems running?  are
you able to roll back to them until you can isolate and resolve your
issues with 5.x?  did you run into these problems on this hardware
before rolling them over to production?  you are positive it is not a
hardware issue?

now that the obligatory questions are out of the way :)  so what was
causing the systems to lock up?  not to doubt your troubleshooting but
I've seen a fair amount of people switch platforms midstream due to
hard to pin down problems only to find it is a (mis)configuration
issue or faulty hardware.

-p


-- 
~~o0OO0o~~
Pete Wright
www.nycbug.org
NYC's *BSD User Group