CPU spikes

Post Reply
doctora
Member
Posts: 93
Joined: 03 Aug 2009 10:12

CPU spikes

Post by doctora » 29 Jan 2015 17:39

I have been receiving snmp traps about cpu crossing above threshold. I am looking for ideas on where to start looking into this. Thanks in advance for any ideas.

User avatar
rekeds
Member
Posts: 103
Joined: 05 Jan 2015 09:22
Location: Riga

Re: CPU spikes

Post by rekeds » 30 Jan 2015 01:21

read logs, monitor utilization, disable debugs?

if l3, is it stable?

devnull
Alcatel Unleashed Certified Guru
Alcatel Unleashed Certified Guru
Posts: 970
Joined: 07 Sep 2010 10:16
Location: Germany

Re: CPU spikes

Post by devnull » 30 Jan 2015 04:24

snmp requests?

All that must be handeld by CPU, can be responsible:
- snmp requests,
- stp handling
- multicast
- ARP handling
- .. ..
if you have longer spikes you can try to analyze in dShell.
>dshell

Working: [Kernel]->spyMax=3
spyMax = 0x55f5c4: value = 3 = 0x3
Working: [Kernel]->
------- SPY TOP TICK USAGE -------
task ticks: taUdldNi 36
task ticks: tShell 13
task ticks: bcmL2X.0 8
idle total: 373 500

------- SPY TOP TICK USAGE -------
task ticks: taUdldNi 22
task ticks: bcmL2X.0 9
task ticks: taVmnCtrl 7
idle total: 377 500

spyMax=0

spyMax = 0x55f5c4: value = 0 = 0x0

User avatar
rekeds
Member
Posts: 103
Joined: 05 Jan 2015 09:22
Location: Riga

Re: CPU spikes

Post by rekeds » 30 Jan 2015 10:32

thanks dev, added your commands to my bookmarks.

doctora
Member
Posts: 93
Joined: 03 Aug 2009 10:12

Re: CPU spikes

Post by doctora » 23 Feb 2015 16:17

Sorry it took me so long to get back to this. The spikes are not service effecting but they bug me.

Thanks for everyones input

rekeds,

read logs - nothing in swlog except me logging in and out,
monitor utilization - nothing strange except the CPU spikes
disable debugs - How do I know if debugs are on?

devnull

I was monitoring the switch with whats up gold. I deleted the device to see if the SNMP requests were contributing. No Change still getting spikes
I am a novice and I dont know what the rest means. I will look some of it up but I probably wont venture into the dshell.

To all that are reading this.

So far I did add the switch to an old version of Omnivista so I could use the statics to time the spikes and then I used port mirroring to catch all the data and pinpoint the time of the spikes. I was hoping to see a bunch of SNMP data but I did not. I am not sure what I should be looking for but I will keep looking and report what I find and how I found it.

It has been running for 208 days. Does anyone think that it could need rebooted? It would tick off some clients but I am in IT so I am fairly used to ticked off clients.

Thanks again for any feedback
Mark

User avatar
rekeds
Member
Posts: 103
Joined: 05 Jan 2015 09:22
Location: Riga

Re: CPU spikes

Post by rekeds » 23 Feb 2015 17:10

we had cpu spikes with the ipv6 long packet bug.
but it was fixed some time ago.

debug ipv6 set ipv6ControlProtocolDisable 0 this helped in my case.

but the cpu 100% was for few hours and wouldn't drop.

are you the only one who can access the switch?
maybe somebody is doing something nasty to it, like scanning, bruteforcing, but that should be in logs.

we have about 80-100 omniswitches, from time to time one of the 6400s goes to 100% and then drops down, but that hasn't bothered me that much as its once a month and I was not able to determine why they do it, we don't use any l3 features also.

matteonasi
Member
Posts: 3
Joined: 03 Oct 2012 05:41

Re: CPU spikes

Post by matteonasi » 01 Sep 2015 04:12

same problem here. all of the 6450 are "soffering" CPU spikes on random bases during working hours. I could only notice that on some device things are fine without vlan port mobility enabled, but I wouldn't say that's the problem. Same config into 6850 devices without any issue.
tried different 6.6.3 and 6.6.4 AOS versions without any difference.
pretty difficult to find out.
any more idea ?

User avatar
thermseeker
Member
Posts: 26
Joined: 08 Jul 2016 08:40

Re: CPU spikes

Post by thermseeker » 24 Jul 2019 10:14

After almost 4 years, any conclusions on this? I keep getting SNMP traps on CPU and Rx/Tx thresholds crossing up, and after a few seconds another trap crossing down.

Post Reply

Return to “OmniSwitch 6850 / 6850E”