Comware

yesterday i enabled traffic collection in automatic mode for all ports in the network via procurve manager.
It caused a spanning tree loop op our core.
Disabling traffic collection and rebooting the core switch(5412zl) solved the problem.
To reproduce the problem i enabled traffic collection on the core and i saw my pings getting worse(with some timeouts). Disabling the collection again solved it.

Does anyone had similar experience with this?
Is there some documentation on how this works?
i didn't saw increased traffic on the network and cpu of switches and firewall was normal.
(i am using procurve manager plus 2.2)

Stieven,

PCM+ 2.2 traffic collection doesn't have any explicit relationship whatsoever to spanning tree. The only configuration performed on devices by traffic collection is configuration of the sFlow MIB for specific ports that sampled data will be acquired from.

If you're seeing STP loops or the switch performance decrease when traffic is enabled, perhaps the switch agent is not handling sFlow well internally with respect to the switch's other duties. You can work around this issue for the time being by manually configuring the ports of the switch for statistics data collection rather than leaving those ports subject to the automatic sampling algorithm. This can be done by selecting the ports of the device and right-clicking to show the configuration options for collection type. Statistics polling will limit traffic's interaction with the switch to reading the port interface counters, with no configuration at all performed.

In the meantime, assuming that you have the most recent release of 5412zl firmware installed, you should open a support call with ProCurve. To my knowledge the only performance problems related to sFlow were resolved almost a year ago, so if problems are still being seen with recent firmware ProCurve is not aware of them.

Regards,

SVB

I can't say explictly whether or not my problem was caused by traffic monitoring, but it did occur not long after having enabled it.

I too have a 5412zl as a core switch, and saw major problems with blades A,C,and D (copper PoE modules) crashing and rebooting and ping timeouts across the network.

Like you we rebooted the switch and things seemed to sort themselves out, but the problem did re-occur (i think later that day). We decided to disconnect all the end nodes and re connect them one at a time, since doing this we have had no problems, but it is quite conceivable that I made some change in PCM without realising it.

I had a similar issue collecting SFLOW stats using PCM+ from a 5308xl. After a period of time, the switch would completely lock and need rebooting.

I reverted to just collecting stats, and everything was ok. I don't have STP enabled on this switch, but it's an interesting coincidence.

Oddly enough, I was collecting SFLOW data from a half dozen other 5308's with the same f/w with no problem. It was just this one switch!

There's an identified bug with some versions of the firmware on 5400s and Sflow which causes a switch crash. We've been informed after having this issue logged that its fixed in release 12.09.

i'm on K11.49. i'm going to upgrade and test again.