Uwe meet Jim, 
Jim, Uwe, Uwe Jim.  :)
Okay, now I'm merging my last messages from both sides here and adding my own 
comments, hopefully out of all this both OpenLab and LTSP can benifit.

On Wednesday, 13 October 2004 16:22, Jim McQuillan wrote:
> A.J.,
>
> interesting info.  I hadn't really looked at what would be causing
> additional NFS traffic, but it makes sense.
>
> If they are booting with PXE, then the initrd image is a separate file,
> and it can be gunzipped and mounted on the loopback device.  Then, the
> /linuxrc script can be edited to add the additional options to the mount
> line.  Once that is done, the filesystem can be unmounted, and the file
> can be gzipped.
This is true, but we use both PXE and Etherboot, and mostly the latter.
>
> I'll add these notes to my list of things for the next kernel package
> release, and I'll give it a try myself.
Thanks for that, this is why I thought I needed to update you with the info 
below.
>
> Thanks,
> Jim McQuillan
> [EMAIL PROTECTED]
>
From me to Uwe earlier.
> Now as for actually implimenting such a change, the difficult way in this
> case is indeed via initrd because the network booting modifies how it's
> store, it's quite tricky to change. 
-- See my mention about using mostly etherboot above.
> However there might be a simpler way.
> Why not edit the rc.S script in the /opt/ltsp tree.
> This probably includes a remount of the nfs share anyway, which one could
> modify.
> Alternatively one should have no problems adding a line like this in it.
> It should be as early on as it can be safely then.
>
> mount -o remount,acregmin=XX[etc] � /

Reply from Uwe later
>Alright, it isn't quite that easy because there is no fstab
That's an odd one, it shouldn't actually NEED fstab, since it's not a normal 
mount. However I think it might be trying to read mtab - and that would fail 
inside the ltsp env.
>but I managed to   
>do that and test it. Unfortunately, it changed nothing. So I was wrong, it is 
>not the cache updates. :-(
Well at least we have that confirmed now.

>Every clients sends one or two nfs query to the server every 5 seconds. That 
>isn't too bad. 
>Actually, it's peanuts. 
Agreed, we get a hundred times more X traffic.
>But then, every 30 seconds, it sends a   
>huge amount of requests. That is very reproduceable if you connect just one 
>client and do a "tcpdump -i eth0 | grep '.nfs'" on the server.
I've done the test on my side, and can confirm this. 
Not that I have any idea what's triggering it.
I can confirm that this is the same regardless of wether /etc/exports is 
written with ro or rw, and ditto for sync or async

>So it's back to square one. :-(
Not quite, at least we ruled something out, and can now look for the cause in 
other places.
-- 
A.J Venter
Lead Developer, DireqLearn 
082 726 5103
http://www.direqlearn.org
http://www.direqlearn.net/olce
http://silentcoder.co.za


-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_____________________________________________________________________
Ltsp-discuss mailing list.   To un-subscribe, or change prefs, goto:
      https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help,   try #ltsp channel on irc.freenode.net

Reply via email to