Summit X460-48p CPU utilization

[EtherNation_Use]

Create Date: Jan 25 2013 4:54AM

Hello All,

I am hoping that someone can shed some light and help me with a problem that I am currently facing in my environment. We recently upgraded our edge switches to Summit X460-48p and deployed it as a stacked solution with each stack consisting of around 4 nodes. Our entire estate including these switches is all running XOS 12.6.3.2.

On these switches and this is not specific to a single stack, I repeatedly see problems with high CPU utilization. Initially (about once or twice) it was the hal and cli processes skyrocketing. However I would probably call these one offs. Whats been killing me is this one process called bcmasync. It continually keeps skyrocketing (70-90%) and has already caused outages wherein a couple of nodes simply rebooted (system watchdog kicking in). A lot of times, the switch stack simply becomes unresponsive. The logs didnt indicate much and Epicenter didnt pickup anything either. As a pre-emptive step, I setup STP based edge port loop prevention, turned off auto polarity on edge ports, set rate limits on broadcast and multicast to rule out other possibilites. Nothing seems to have helped.

I do have a case with Extreme and they weren't initially sure either. I have already provided logs, tech support etc to no avail. They have finally come back asking for a debug to be done which I am planning. At this stage I am not sure whats happening and I almost had another outage yesterday. Any thoughts on this would be really appreciated and sorry for the long email. Tks again.

(from Anush_Santhanam)

[Bill_Stritzinge]

Just saw this post, hope you had it figured out but just in case you are still seeing the issue... You need to upgrade the code to at least 15.1.x and bootrom 2.0.7 or later. We did some major revisions to the CPU usage in this combination and it will also give you about 30% more CPU cycles in normal situations.

[Jim_Keeffe]

What was the result of you switching off port speed 100Mb?

[EtherNation_Use]

Create Date: Feb 12 2013 6:29PM

I've seen this issue (although not to the degree of the OP) as recently as XOS 15.2.2.7-patch1-6. We don't run RSTP in our environment and don't seem to see any port flapping, either. The one thing that separates the affected switch from our unaffected X460s is that we have certain ports configured only to 100Mb. While not link flapping, this does seem to be link-related; I'm going to try switching this off after hours and report back on my results. (from Ansley_Barnes)

[EtherNation_Use]

Create Date: Jan 25 2013 9:02PM

Hello Excalibur

I did some checking and from the internal information it seems that this is seen with a lot of port flapping and seems to be more of an issue with RSTP.

I think TAC is the best way to go as they can narrow it down. In the reports I saw we may need to upgrade code but that is something that TAC should recommend.

Hope that helps.

P (from Paul_Russo)