Re: Onbar and Veritas Netbackup?
Posted in 2003
Hi
The BAR_TIMEOUT (hardcoded to be 600 seconds) is used to set the maximum
time onbar will wait for a XBSA API call.
First a flag is set to indicate that a XBSA api call is about to take
place.
Then a XBSA api call is invoked.
Then the flag is clear to indicate that the XBSA api call has completed.
In the meantime, the onbar signal handle has been setup to handle SIGALRM
(alarm clock). Inside that SIGALRM handler, if a XBSA api call is in
progress and the BAR_TIMEOUT has been exceeded, then the warning is given.
At this point, the timeout value is effectively increased and onbar
continue to wait for the completion of the XBSA api call.
thanks. davis.
briceavila@hotmai
l.com (Brice To: informix-list@iiug.org
Avila) cc:
Sent by: Subject: Re: Onbar and Veritas Netbackup?
owner-informix-li
st@iiug.org
08/08/2003 07:18
AM
Please respond to
briceavila
I don't believe that BAR_TIMEOUT is causing the performance problem,
but the other way around: a performance problem is causing the
BAR_TIMEOUT warning.
I'd like to learn more about this parameter also. I understand that
this is hard-coded to 600 seconds (10 minutes). There is an Informix
bug 115147 that stopped a backup if a large dbspace reached 10 minutes
before being completely scanned/backed-up, and perhaps the fix is to
change this to a warning? This bug I believe is fixed in:
8.31.UC1
8.40.UC1
9.30.UC1
Anyone know more about BAR_TIMEOUT? Is it configurable? If so, how?
TIA.
Brice Avila
Minneapolis, Minnesota
"Venkatesh Konnur" <VNKONNUR@kcc.usda.gov> wrote in message
news:<bgtj7k$1ln$1@terabinaries.xmission.com>...
> Hi,
> we use Netbackup version 4.5 with IDS 9.4.UC1 version on Sun OS 5.8.
>
> In the bar_act.log, we very frequently see the following warning:
>
> 2003-08-07 04:34:18 16762 16760 Warning: BAR_TIMEOUT Storage Manager
Progress
> may be stalled.
>
> Does anybody know what the above means...How could this be
recitified...Becuase of the above out level-1 and level-0 backups are
taking longer than it should be...
>
> Thanks
> Venaktesh
>
>
> sending to informix-list
sending to informix-list