Parallel job hanging on oracle SQLloader ?

siauchun84 · Post by **siauchun84** » Wed Sep 29, 2010 9:26 pm

When your job hang, have you checked on the task manager in the server on the total numbers of the Osh.exe? I have faced similiar problem before which there are lots of Osh.exe processes still exist in the task.

wbeitler · Post by **wbeitler** » Wed Sep 29, 2010 11:24 pm

Thanks for that. Weird problems make you look for weird things...
I'll keep you posted.

wbeitler · Post by **wbeitler** » Thu Sep 30, 2010 3:59 am

siauchun84,

When your job hang, have you checked on the task manager in the server on the total numbers of the Osh.exe?

Aborting the job manually actually did get rid of 12 osh.exe processes.
Care to join how you solved your problem ?

siauchun84 · Post by **siauchun84** » Sun Oct 03, 2010 10:07 pm

wbeitler wrote:siauchun84,
Aborting the job manually actually did get rid of 12 osh.exe processes.
Care to join how you solved your problem ?

Hi, wbeitler, what I did was created a vbscript batch file to kill all the osh.exe before I trigger my next job seem I was using Windows platform.
*Note: Killing the osh.exe will causing all the running jobs terminated.

You may copy the following into notepad and save as .vbs in the server (Just double click it if you have direct access to the server. If not, create a seq job to trigger the script):
-------- VB Script --------
Dim wql
Dim wmi
Dim oResults

Set wmi = GetObject("winmgmts:")
wql = "SELECT * FROM win32_process WHERE name ='osh.exe'"
Set oResults = wmi.ExecQuery(wql)
For Each Process In oResults
Process.Terminate
Next
set wmi = nothing
---- End of VB Script ----------

wbeitler · Post by **wbeitler** » Wed Oct 06, 2010 5:58 am

Siachun,

thanks for sharing, but no option here since we're running multiple jobs in parallel.

Jobs actually didn't crash after we've enabled the APT_RECORD_COUNTS reporting environment variable. That somehow seems to 'keep the communication alive' ?! Would that make any sense ?!

wbeitler · Post by **wbeitler** » Thu Oct 07, 2010 3:41 am

Doesn't make any sense... Job hanging again, although we now do get the second 'Load complete' message, but the 'Record count' for only 1 of the nodes... Waiting in vain for the second node to return it's record count

chulett · Post by **chulett** » Thu Oct 07, 2010 5:29 am

So... did you ever bring support into the picture? Have they tried to figure out what in the heck might be going on? Seems like you are well and truly into their terrritory now.

wbeitler · Post by **wbeitler** » Thu Oct 07, 2010 5:50 am

PMR- logged. Still waiting for their answer.

wbeitler · Post by **wbeitler** » Wed Nov 17, 2010 8:03 am

Installed patch JR36567 and Fixpack 3 as adviced by IBM Support.
Didn't fix the problem though... Any new thoughts?

William