Mailing List Archive

Random "pausing" on ICX6450
I'm scratching my head with an issue, I was wondering if anyone else has
seen it?

At various points throughout the day, I've noticed the ICX "pause" -- that
is, not transfer any traffic. As far as I can tell, this is not a case of
it not forwarding traffic, but not emptying the buffers.

Things have much improved since I stopped RANCID from logging in every
hour, but I'm still seeing the occasional pause of several hundred
milliseconds. Has anyone else seen this?

Any input greatly appreciated.

Matthew Walster
Re: Random "pausing" on ICX6450 [ In reply to ]
On 23 January 2014 11:55, Jeroen Wunnink | Atrato IP Networks <
jeroen.wunnink@atrato.com> wrote:

> Check for excessive SNMP polling, Observium has a bad habit of pulling
> entire MAC tables from switches by default, locking the CPU at 100% while
> doing it if it's a big table
>

An interesting theory -- I believe we have OpenNMS which may try and
re-auto-discover that switch occasionally.

Thanks for the idea!

Matthew

​​
Re: Random "pausing" on ICX6450 [ In reply to ]
On 23/01/2014 14:46, Matthew Walster wrote:
> An interesting theory -- I believe we have OpenNMS which may try and
> re-auto-discover that switch occasionally.

it's worth bearing in mind that ironware is a non-preemptively multitasking
operating system which doesn't support process prioritisation. This means
that important stuff (e.g. lacp) is treated the same as junk (NMSs polling
mac tables) by the switch processor.

Obviously this shouldn't affect the forwarding plane of the device. If it
does, there is a problem which needs to be escalated to the brocade tac.

Nick

_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp