[iyonix-support] Re: Interrupt failure in EtherK module (revisited)

  • From: Gary Locock <gary@xxxxxxxxxxxx>
  • To: iyonix-support@xxxxxxxxxxxxx
  • Date: Mon, 05 Feb 2007 20:03:11 +0000 (GMT)

In article <59fa53b04e.rik-mail@xxxxxxxxxxxxxxx>,
   Rik Griffin <rik.griffin@xxxxxxxxxxxxxxxxxxx> wrote:
> In message <4eb019f502gary@xxxxxxxxxxxx>
>           Gary Locock <gary@xxxxxxxxxxxx> wrote:

> > In article <f1bbcaae4e.rik-mail@xxxxxxxxxxxxxxx>,
> >    Rik Griffin <rik.griffin@xxxxxxxxxxxxxxxxxxx> wrote:
> > > About a year (almost exactly) I discovered a problem with the EtherK
> > > module 0.18 that caused the IRQ in use by the network device to fail
> > > after a period of intensive use. I think Gary Locock found the same
> > > problem as a result of heavy use of ShareFS.
> > 
> > > Castle tracked down the problem to the EtherK module, and version 0.19
> > > was released which seemed to cure the problem
> > 
> > > So, has anyone else seen this problem recently?
> > 
> > Hmmm.  Interesting.  Big file copies have been reasonably reliable until
> > recently, but I had the server stiffed on me a couple of times only last
> > week. Not confined to ShareFS; one was with LanMan98, albeit after a
> > week's uptime and fairly heavy ShareFS traffic. I can't be as specific as
> > you are about the details.  However I would say that 0.19 was a big
> > improvement rather than a complete cure for the Iyo's networking issues.

> I set up a test machine with RO5.12 and left it running over the weekend,
> no problems when I checked it this morning.

My impression is that background network activity doesn't give us the degree
of failure levels that you have reported in the past.  In our case there is
always a surge of network (and/or copying) activity on the host itself that
gives rise to a failure.  (Our network is reasonably busy, but I think,
probably operating at only a small fraction of ultimate capacity.)

> Which doesn't prove anything of course, but maybe there's an interaction
> between some other OS component and EtherK.

Indeed; and in our case it couldn't be limited to just OS components.  I
doubt if it is anything particularly recent either.  'Big file' copies have
always had the potential to freeze up the Iyo, and of course, once frozen,
there is no realistic way of finding where the problem might have been. It
has been a /lot/ less troublesome than it once was, pre-0.19, but the
occasional episode just serves to show that the problem hasn't completely
gone away.

Sadly, RO networking is not in the same league of bulletproof reliability as
Mac OS X and Linux, to name but two; which, considering the common
ancestries, is a pity.

-- 
Gary Locock, Network Manager, Bablake Junior School
Coundon Road, Coventry CV1 4AU
School Website: http://www.bablakejs.co.uk
Private mail:  g a r y (at) l o c o c k . c o . u k

---
To alter your preferences or leave the group, 
visit //www.freelists.org/list/iyonix-support
Other info via //www.freelists.org/webpage/iyonix-support

Other related posts: