Excessive interrupt disablement time

[strikelit]

I am running aix V5.3 and getting the following error message...."Excessive interrupt disablement time "

I have never seen this and I cant tell what is causing this.

ABEL: DELAYED_INTS
IDENTIFIER: A2205861

Date/Time: Tue Aug 29 21:18:46 CUT 2006
Sequence Number: 6052
Machine Id: 000C758A4C00
Node Id: chiprod
Class: S
Type: PERF
Resource Name: SYSPROC

Description
Excessive interrupt disablement time

Probable Causes
SOFTWARE PROGRAM

Failure Causes
SOFTWARE PROGRAM

Recommended Actions
REPORT DETAILED DATA

Detail Data

TICKS 51 (3) decr, IAR 2A2858 [2A2858], LR 108860 write_relp+1B4
[2A2858]
write_relp+1B0
v_flush+33C
v_preflush+1B4
[2A247C]

 

[cspilman]

Maybe this link helps?

http://www-1.ibm.com/support/docview.wss?uid=isg1IY77637

Regards,
Chuck

 

[rick26]

I have the error on one of our systems also. IBM rpeorts another fix IY78131, on

http://www-1.ibm.com/support/docview.wss?uid=isg1IY78131

However the specified APAR is already appl;ied on the system with AIX 5300-04-CSP.

 

[DukeSSD]

Don't worry about them unless the system becomes unstable - if it does raise a Sev1 PMR as usual.

In 5.3 they introduces a load of interrupt checking, to catch interrupts that went unanswered and either highlight potential problems or help performance, but in the process it has highlighted loads of interrupts that have been going unanswered, now they will spend a few months / TLs sorting out all the device drivers and timing problems that cause them to happen.

I expect there will be a few APARS but they will all be built into the next few TLs so ignore them if you can - or go chasing after those APARs.

 

[rick26]

IBM Support confirms that fileset bos.mp64 at level 5.3.0.40 should solve the problem, but they also mention 5.3.0.50.
IBM lab has been asked to deliver an e-fix (untill we upgrade to 5.3 TL05).

 

[DukeSSD]

http://www-1.ibm.com/support/docview.wss?uid=isg3T1000678

explains what is going on.
and how to disable checking - but that is just masking the problem - not that is a real problem..
Seems to be CPU interrupt disablement checking and has been happening for ages, it is just that now AIX is telling us about it.
I heard fixes are being worked on but it may be some time before they are all available.

 

[rick26]

Thanks for the reaction. Today I was in contact with IBM lab: they are working on fixes but they advised me to ignore the errors as long as there is no significant performance reduction.

 

[kHz]

IBM Support confirms that fileset bos.mp64 at level 5.3.0.40 should solve the problem

It doesn't.

 

[DukeSSD]

khz,
there are all sorts of processes that can cause these errors.
if you look at the detail data section you will find these errors can be caused by all sorts of functions.
there will be LOADS of filesets needed to fix all of the issues in AIX let alone other apps that use interrupt dissablement.
unless you think they are related to a real performance problem, don't worry about them.
the "problem" has always been there, it is just that the later TL is actually logging them now.
the reason is to allow IBM to start designing out some of these delays to help further improve performance.