AdvfsIORetryControl Sysconfig paremeter

[SAKET_5]

Hi All,

Does anyone know the specifics on this sysconfig parameter for AdvFS. I understand that a value of 0-9 can be assigned to this parameter, where 0=no retries and 1-9 means the number of retries.
We recently noticed that during an array controller software upgrade (ACS on a HSG80 based storage array)in an online fashion - no host downtime - upgrade carried on one controller at a time -> the only Tru64 Unix cluster containing two members (running Tru64 Unix V5.1B-PK4) experienced a few I/O errors and domain panics on several Advfs domains. In the end, the panics seemed to have caused a crash on both members.

The console log message indicated that the current value of "AdvfsIORetryControl" paramter on the system is set to 0 - no retries and that had it been set to a higher value the I/O operation would have succeeded.

I just wish to gather some specifics on this paramter, what do the values 1-9 actually mean ,i.e if the value is set to 5 does it mean Advfs will retry an I/O in the event of it not succeeding in the first attempt 5 times at a time interval of ?? seconds/minutes??

Thanks in advance.

[Uwe Zessin]

Found some information in the BC-EVA V2.3 Admin Guide (yes, I know it happened on a HSG):

http://search.hp.com/redirect.html?type=REG&qt=AdvfsIORetryControl&url=http%3A//h20000.www2.hp.com/bc/docs/support/SupportManual/c00215123/c00215123.pdf&pos=5

[SAKET_5]

Uwe,

Thanks for the link.

Although it provided an extra piece of information that I wasn't aware of, i.e. "HP recommends setting the value of this parameter to 1" - the information presented in this guide is less than sufficient to actually figure out what is the time interval between each I/O Retry interval for particular values of this parameters - is it in minutes/seconds?

Also, although I don't know the details on this parameter as yet, I find it a little strange that the default value of this parameter is set to 0.

Anyway, I have logged a call with HP - hope its not months before I get to speak with someone who knows the specifics of the topic.

Have you had any similar experience with Rolling Upgrade of ACS?

Regards,
Saket.

[Uwe Zessin]

Hello Saket,

I did a rolling upgrade, but the servers were down during the task - everything went fine afterwards. If I recall correctly (it's 1+ year ago) my colleague was doing maintenance to the servers (we try to overlap our activities).