Citrix Netscaler Fragmentation Issue

[George Bonser]

> -----Original Message-----
> From: Jan Dirnberger [mailto:jpd at space.net]
> Sent: Thursday, June 09, 2011 7:26 AM
> To: George Bonser
> Cc: IPv6-OPS
> Subject: Re: Citrix Netscaler Fragmentation Issue
> 
> Hi,
> 
> Am 08.06.2011 um 18:26 schrieb George Bonser:
> > There is no released code to fix this issue at this time.  Bottom
> line
> > is that if you put an IPv6 VIP on a Citrix Netscaler, you risk
taking
> > your entire load balancer out of service.
> 
> 
> do you have a Bug-ID or sth? I'd like to bash a little on our contact
> persons at Citrix....
> 
> Jan Dirnberger

I had a bug ID that they believed it was, but this is appearing more
like a generic bug than a specific IPv6 bug, we are not passing any IPv6
at all, have IPv6 disabled, and continue to have problems with one of
the units.  When it gets in this state you will see messages in
/var/log/messages like:

Jun  9 04:42:17 <local0.warn> sjc-lb-02 nsconmsg: IOCTL failed due to
Failed to read data from Aggregator
Jun  9 04:42:37 <local0.warn> sjc-lb-02 nscollect: IOCTL failed due to
Failed to read data from Aggregator

Then you start seeing evidence of it flapping between primary and
secondary:

Jun  9 04:02:09 <local0.notice> 10.247.224.212 06/09/2011:12:02:09 GMT
sjc-lb-02 PPE-0 : EVENT STATECHANGE 2841 :  Device "self node
10.247.224.212" - State Primary

Jun  9 04:02:09 <local0.notice> 10.247.224.212 06/09/2011:12:02:09 GMT
sjc-lb-02 PPE-0 : EVENT STATECHANGE 2843 :  Device "self node
10.247.224.212" - State Secondary

It looks like it is load related to me.  So it starts out with a failure
to read data from Aggregator which appears to be a process that collects
stats from the packet engines. Then you get floods of messages about
services to the log, then you start seeing it flap between primary and
secondary, and finally it just goes deaf on the network and you have to
get in by console and kick it.

It almost behaves like it is a problem with the packet engines providing
more stats than the Aggregator can swallow, it starts packing up, then
maybe the packet engines run out of buffer or something and they start
to blow up and then it eventually goes deaf.  I am not saying that is
*the* problem, but that is how it behaves.

The had a bug ID but the latest word is that they believe this might be
a new one and could be NIC related.  To me it just looks like the
Aggregator can't get out of its own way.

This is running NS9.2-50.4