NBU Env Freezes up....

[jdespres]

I help manage 5 NBU 4.5 fp6 environments......

All masters are Sun systems.....

Of these I have seen the environment freeze up on 3 of them!

Of course these 3 are the largest....

Here are the symptoms:

1. Active jobs complete but do not finish. Jobs say they have completed. But don't leave the state of Active.

2. Queued jobs do not go Active...

3. New jobs do not Queue up....

4. Jobs manually started do not get started....

Has anyone seen this issue?

 

[ugreentrailgirl]

I've seen this behavior before. It's a feature of bpsched for this version. The fix is either to get the "special" bpsched patch or upgrade to fp7. I applied the patch.

 

[jdespres]

Thanks for the input......

Veritas gave the following:

1. Adjust tcp_time_wait_interval to 10,000...

2. upgrade the NBU software...

3. upgrade the hardware....

 

[jdespres]

Well......

Veritas has told me....

There is no patch....

Bummer!

 

[ugreentrailgirl]

Sorry to hear that but I wonder if it's still on their ftp site. If you want to search I think the path is NB_45_6_F_S1021.solaris_265048.tar

 

[jdespres]

Bummer..........

That particular patch is already there....

Thanks for the effort....

 

[DRFranco]

I am having the exact same issue in on a Windows 2003 master server with NetBckup 5.1 MP3. bpsched process goes to 100% utilization and no new jobs will run. Jobs complete but stay in the active state. I have seen this issue since 4.5FP6.

 

[jdespres]

Check your "CLIENT_CONNECT_TIMEOUT" value....
the default is 300....

 

[DRFranco]

What is the best setting for "CLIENT_CONNECT_TIMEOUT" ?

 

[DRFranco]

Hmmm.. I found this for the UNIX environments. It may be usefull. It explains in depth how the scheduler works and what could cause it to hang.

http://support.veritas.com/docs/274559

 

[creakyjoe]

@jdespres - ugreentrailgirl is correct about the NB_45_6_F_S1021.solaris_265048.tar being the patch, at the very least that patch addresses the bizarre issue of bpsched core-dumping all over the place.

We've since gone FP8 a couple months ago as have suffered non-stop write/position errors since we went FP6 3yrs ago...

As to CLIENT_CONNECT_TIMEOUT, i've never found a need to alter from default. However i did alter the following 3 params in bp.conf (on Master & ALL clients) some years ago, may have been back in the 3.4 or 3.11 days -

# make sure the next 3 timeout params are identical..
# - bpend_timeout, bpstart_timeout & client_read_timeout...
BPEND_TIMEOUT = 99999
BPSTART_TIMEOUT = 99999
CLIENT_READ_TIMEOUT = 99999

Hope that helps

Rich