[geeks] Well, THAT was a setback

[Phil Stracchino]

I've heard people talk about the Linux TCP stack falling down under
heavy load, but never had it happen to me.  Until today.

I just got done doing a totally clean OS reinstall on babylon5, my main
workstation, on a mirrored pair of SATA disks on a MegaRAID-based LSI
Logic battery-backed SATA-RAID controller.  The old OS was a custom
Linux install based originally on Slackware 7, running a 2.4.37 kernel.
 The new is Gentoo 10.0 on a 2.6.32 kernel.

With the current version of Bacula installed, I set up a storage daemon
pointed at my new-to-me LTO2 drive, which is now the only SCSI device
and this has an Adaptec 29160 controller all to itself.  The drive syncs
up at the full 160MB/s.  I started a full backup of a single client and
let it run for a few minutes ...  looked good, so I started four or five
more clients.  And about two minutes later, the network stack fell over.


I have never, ever seen that happen before...  I guess I won't be using
my LTO2 drive attached this machine *at all*.

This does raise the question of what I AM going to attach it to...  I
can't put a SCSI card into the main server and attach it to that,
because I wouldn't risk operating an LTO2 drive in the environment where
the rack is located.  It gets too cold.


-- 
  Phil Stracchino, CDK#2     DoD#299792458     ICBM: 43.5607, -71.355
  alaric at caerllewys.net   alaric at metrocast.net   phil at co.ordinate.org
         Renaissance Man, Unix ronin, Perl hacker, Free Stater
                 It's not the years, it's the mileage.