[j-nsp] KB21404 / MQchip 0 XE 0 Throttle

[Paul Stewart]

Has anyone got caught by this PR?  We just deployed a pair of MX10's for a
customer.  One of them (with the greater amount of traffic so far) is
showing this in the logs:

 

Feb  9 14:03:14  dis1.beachburg1 tfeb0 MQchip 0 XE 0 Throttle: Last 10
seconds have seen interrupt throttling at least once per second

Feb  9 14:04:15  dis1.beachburg1 tfeb0 MQchip 0 XE 0 Throttle: Last 19
seconds have seen interrupt throttling at least once per second

Feb  9 14:05:26  dis1.beachburg1 tfeb0 MQchip 0 XE 0 Throttle: Last 10
seconds have seen interrupt throttling at least once per second

Feb  9 14:06:35  dis1.beachburg1 tfeb0 MQchip 0 XE 0 Throttle: Last 10
seconds have seen interrupt throttling at least once per second

Feb  9 14:07:37  dis1.beachburg1 tfeb0 MQchip 0 XE 0 Throttle: Last 10
seconds have seen interrupt throttling at least once per second

 

We first seen this with 11.4R1.14 release and have since downgraded to
11.2S4.1 release and it's still occurring.  It is supposed to be fixed in
11.2R2 release and I've been waiting another full day for JTAC to get back
to me.

 

I'm trying to find out if this was fixed in the 11.2S4.1 release or not . we
are using these for BRAS functions and I'm told to stick to 11.2S code if
possible (or bleeding edge 11.4R1.14 where this is a known problem I'm
told).

 

The effects are basically we're seeing about 1 packet in every 1000 dropped.
When the drop occurs, the log file records an entry as per above.  Our fear
is that if we put more traffic on these boxes, the problem could compound
and create much greater packet loss (or other issues not seen yet).

 

I am loading 11.2R5.4 release onto these boxes as we speak to see if there's
any change in behavior.  Unfortunately this release doesn't have one of the
BRAS features we had hoped but can work around that if it means stable
performance on the box.

 

Any suggestions/experiences are most welcome.

 

Paul