PDA

View Full Version : SmartView Monitor showing mgmt console at 100% CPU - R65



dougg01
2009-10-26, 09:39
One of our managment consoles (is in mgmt HA) is showing 100% CPU util on the Montior GUI and shows 3.64GB of active virtual memory.

This is on a Dell 2950 dual-quad core w/ 8GB of RAM running on SPLAT.

On the box, the average is way below this.

Odd thing is top always shows 698-699% idle (cumulative across all 8 procs)and total util across them all seems to never exceed 102%. The system does not seem to hammer any particular processor, but does want to revisit cup06 more than the others.

Another mgmt console (not yet in mgmt HA) typically shows <20% util and the command-line reflects this. Identical hardware and code level. Load on this one is about 1/3 to 1/2 LESS than the 1st one.

Seems odd.

Anyone else seen this....?

daz306td
2009-10-26, 12:40
I have seen this within my own environment - there was a known hotfix, but it didnt need to apply it as it seemed to oddily cure itself. Sorry I dont have any specfics on the hotfix but I would suggest contacting your support team.

Out of interest what build of 2.6 are you using? You can find this out by running 'ver' at the command line

melipla
2009-10-26, 13:57
there was a known hotfix

Yes, there are several hotfixes, from R65 - R70.1, each with their own, non included in an HFA that I'm aware of.


Odd thing is top always shows 698-699% idle (cumulative across all 8 procs)and total util across them all seems to never exceed 102%. The system does not seem to hammer any particular processor, but does want to revisit cup06 more than the others.

Another mgmt console (not yet in mgmt HA) typically shows <20% util and the command-line reflects this. Identical hardware and code level. Load on this one is about 1/3 to 1/2 LESS than the 1st one.

Seems odd.

Anyone else seen this....?

I would agree, very odd. If you do a "H" in top it should show you threads. It would help to know which process is using all of the CPU.

Then try rebooting to see if it fixes it...

whatchamccallum
2009-10-27, 09:38
If you are running 2.6 there is an bug in SmartView monitor. The 2.6 kernel changed the datatype of a variable in /proc/stat

The new hotfix just fixes the datatype of the variable that the value is read into.

If you run uname -a it will show you if you are running 2.4 or 2.6. If it is 2.6 then see sk36634

dougg01
2009-10-27, 15:06
I'm slow to reply. I found the SK and requested the hotfix. I applied it last night.

Symptoms: top shows 99-102% busy (698% idle cumulative across 8 procs) and are constant. Load moves across the processors, but never varies outside this range. Monitor maintains a constant 100% util.

After the hotfix: top shows 1-5% busy most of the time. Policy push, mgmt sync, etc cause a momentary upswing. Not much. Monitor shows nearly constant 0-1% utilization most fo the time.

Goodness.

I asked and the library file used in the hotfix is one created during the build, but not utilized when it was rolled out. The hotfix uses a now-modified version of the file which is why it appears to be of an earlier origin than the current.

kaydo
2009-11-06, 11:56
To anyone who has installed this hotfix, can you please tell me if it requires a reboot or causes any type of service interruption at all? I received the hotfix but it didn't come with any instructions.

Thanks.

melipla
2010-01-07, 12:43
In case anyone was wondering, this hotfix isn't included in R70.20 and it hasn't been ported yet.

lammbo
2010-01-07, 13:27
Great... So:

1) We've been posting/reporting this bug for almost 2 years
2) CP has the code to correct it from when they developed the product (that's why the fix to this are files older than production code)
3) They still can't seem to get this into any HFA for R65 or R70 and we've had what?!? R65 HFA_30, 40 & 50 as well as R70 HFA_10 & 20 = 5 HFAs to correct this.

In the Army, we would have said they have their heads in rectum defilade. It just can't possibly be that hard to include this in an HFA and resolve this without a one-off Hotfix.

HFA Fix > Added support calls

serlud
2010-03-06, 10:52
Great... So:

1) We've been posting/reporting this bug for almost 2 years
HFA Fix > Added support calls

We have the same issue with R62,R65,R70, R70,1 -CP will still improve this..

CP should also represent an good works, that is why you will see 1000* SK with following sollutions : Please call as and we provide you with hotfix for this issue.
This Service request will be resolved in a second and it help CP to provide customer with very *good* statistic how long does it take to resolve an issue.

In other case CP just close old SR (too long one week for high/high), and open an new one ......just to make sure the customer will get support for payed supports...

The biggest problem they do not provide any SK for some known issues at all (probable only for internal CP SK) >> that is why all customer now or later should open an SR for already known but still not resolved issue.