multiload utility in Parallel mode

nivas · Post by **nivas** » Mon Jan 10, 2005 2:19 pm

Hi

I am trying to use Multiload Plugin in Parallel mode. The job is fine in sequential mode. But when I set the mode of execution in parallel mode the job is aborting. I have 4 four nodes.
THe job is aborting . IN the .rpt file the error message I am getting is CKP tble already exists and returing error code 12. I think the application is trying to create 4 CKP tables because it has 4 nodes. So When it is creating second CKP table the job is aborting. May be some where I need to specify something so that it will create 4 CKP tables with different names or the application will create 4 CKP tables in four nodes.
Could anybody suggest.
thanks
Nivas

T42 · Post by **T42** » Wed Jan 19, 2005 5:40 pm

Can you please paste the log, particularly the error message, and the warning messages just before that error messages?

Thanks.

nivas · Post by **nivas** » Thu Jan 20, 2005 9:46 am

Hi

The following is the error message from .rpt file created by MulitLoad Utility.
-------------------------------------------------------------------------------------
0001 .logtable ua_edw_devl.CKP_CLAIM_FS_PX1;
0002 .logon kansas/svundi,;
**** 10:10:27 UTY8400 Default character set: ASCII
**** 10:10:27 UTY8400 Maximum supported buffer size: 64K
**** 10:10:28 UTY1008 RDBMS failure: 3803, Table 'CKP_CLAIM_FS_PX1' already exis
ts.
**** 10:10:28 UTY6215 The restart log table has NOT been dropped.
**** 10:10:29 UTY2410 Total processor time used = '0.02 Seconds'
. Start : 10:10:27 - THU JAN 20, 2005
. End : 10:10:29 - THU JAN 20, 2005
. Highest return code encountered = '12'.
After this it is writting Processing MultiLoad Phase Messages
Then it is writting MultiLoad Initial Phase Messages
Then it is writting MultiLoad DML Transaction Phase Messages
Then It is not writting any messages from Multiload Acquisition Phase onwards
----------------------------------------------------------------------------------

The following are the error messages from Job log
*****************************************************
Info: CLAIM_FS_IN,3: Info: CLAIM_FS_IN: MultiLoad process 18028 has started
Info: CLAIM_FS_IN,0: Info: CLAIM_FS_IN: MultiLoad process 18031 has started
Info: CLAIM_FS_IN,2: Info: CLAIM_FS_IN: MultiLoad process 18032 has started
Info: bldClaimFsInSRI(trn).#0.trn: DSD.StageRun Active stage starting, tracemode = 0.
Info: bldClaimFsInSRI(trn).#2.trn: DSD.StageRun Active stage starting, tracemode = 0.
Info: bldClaimFsInSRI(trn).#1.trn: DSD.StageRun Active stage starting, tracemode = 0.
Info: bldClaimFsInSRI(trn).#3.trn: DSD.StageRun Active stage starting, tracemode = 0.
Info: CLAIM_FS_IN,1: Info: CLAIM_FS_IN: MultiLoad process 18034 has started
Warning: CLAIM_FS_IN,1: /bin/echo: Broken pipe
Warning: CLAIM_FS_IN,0: /bin/echo: Broken pipe
Fatal: CLAIM_FS_IN,0: Operator terminated abnormally: received signal SIGSEGV
Warning: CLAIM_FS_IN,2: /bin/echo: Broken pipe
Fatal: CLAIM_FS_IN,2: Operator terminated abnormally: received signal SIGSEGV
Fatal: CLAIM_FS_IN,1: Operator terminated abnormally: received signal SIGSEGV
Fatal: main_program: Step execution finished with status = FAILED.
Info: main_program: Startup Time 0:10 Production Run Time 1:00
Info: Contents of phantom output file =>DataStage Job 403 Phantom 17952OpenSocket() returned 7
DataStage Phantom Finished.
Control: Job bldClaimFsInSRI aborted.
Fatal: bldClaimFsInSRI.#3.trn.src-Input.src: ds_ipcgetnext - timeout waiting for mutex
Warning: Attempting to Cleanup after ABORT raised in stage bldClaimFsInSRI(trn).#3.trn
Warning: Abnormal termination of stage bldClaimFsInSRI(trn).#3.trn detected
Control: Job bldClaimFsInSRI(trn).#3 aborted.
Fatal: bldClaimFsInSRI.#2.trn.src-Input.src: ds_ipcgetnext - timeout waiting for mutex
Warning: Attempting to Cleanup after ABORT raised in stage bldClaimFsInSRI(trn).#2.trn
Fatal: bldClaimFsInSRI.#1.trn.src-Input.src: ds_ipcgetnext - timeout waiting for mutex
Warning: Attempting to Cleanup after ABORT raised in stage bldClaimFsInSRI(trn).#1.trn
Fatal: bldClaimFsInSRI.#0.trn.src-Input.src: ds_ipcgetnext - timeout waiting for mutex
Warning: Attempting to Cleanup after ABORT raised in stage bldClaimFsInSRI(trn).#0.trn
Warning: Abnormal termination of stage bldClaimFsInSRI(trn).#2.trn detected
*********************************************************
Hope the above information is helpful for investigation.
thanks
Nivas

T42 · Post by **T42** » Thu Jan 20, 2005 1:49 pm

Fix this error:

**** 10:10:28 UTY1008 RDBMS failure: 3803, Table 'CKP_CLAIM_FS_PX1' already exis

* * *

Apparently, you need to drop this table.

nivas · Post by **nivas** » Thu Jan 20, 2005 3:18 pm

Hi

There are no work tables or log tables or error tables before running the job. THe multiload also released on target table. Basically I am dropping all support tables and target table.I am recreating the target table before running the job.

thanks
Nivas

T42 · Post by **T42** » Thu Jan 20, 2005 4:30 pm

Well, apparently, it is still there.

Did you ensure that the metadata for your database are fully updated based on this?

Try a different table name -- something random. Does the error still happen?

nivas · Post by **nivas** » Thu Jan 20, 2005 5:04 pm

Hi

I did that couple of times by creating new tables. The job is working fine in parallel mode with single node. If I put more that one node in the node constraint mapping I am getting this error.

thanks
Nivas

T42 wrote: Try a different table name -- something random. Does the error still happen?

T42 · Post by **T42** » Thu Jan 20, 2005 9:59 pm

Aha! Fascinating. Here's how EE works:

They spawn multiple processes, one set per node. Apparently, this job is incapable of determining which node is responsible for creating the table, so therefore, other nodes are attempting to create it, and thus failing, ran home crying to its momma.

Unfortunately, anything short of a patch or upgrades will not fix this behavior for you, unless there's a hidden option (which I can not find out, since I don't have a copy here with me at the moment.)

Sorry.

ray.wurlod · Post by **ray.wurlod** » Thu Jan 20, 2005 11:13 pm

I was thinking along those lines - multiple processing nodes all attempting to create the same table.

The obvious solution in a parallel environment is to have a serial environment (you or the DBA) look after single-threaded processes, such as CREATE TABLE, and have the parallel streams then process the data into the table.

l_homme_de_rio · Post by **l_homme_de_rio** » Fri Jan 21, 2005 5:59 am

Hi,
I guess You're trying to load a terradata tables isn't it ?? If so did you set "-sessionsperplayer" and "-requestedsessions " as DB options ???
All the best

nivas · Post by **nivas** » Fri Jan 21, 2005 9:47 am

Hi

My Target Database is Teradata and I am using MultiLoad Plugin for loading in parallel mode.
Could you please tell me where and how do I set these DBOptions.

thanks
NIvas

l_homme_de_rio wrote:Hi,
I guess You're trying to load a terradata tables isn't it ?? If so did you set "-sessionsperplayer" and "-requestedsessions " as DB options ???
All the best

nrevezzo · Post by **nrevezzo** » Mon Jan 24, 2005 2:38 pm

You need to discuss multiload functionality with your DBA.
You can not run multiple multiloads against the same table concurrently.

baabi_26 · Post by **baabi_26** » Mon Jan 24, 2005 5:57 pm

Hi,

The reason why you are facing this problem is already explained in previous posts. This might help you.

1. Change the Execution mode of Multiload stage to Sequential mode.

2. Drop all the log table, error tables and work tables that you have specified.

3. On the main table you are trying to multiload do this

"Release Mload <table_name>"

nivas · Post by **nivas** » Tue Jan 25, 2005 11:41 am

Hi Babi

I am fine with sequential mode or parallel mode with single node. Since we have 4 nodes available I am trying to run the job on 4 nodes. As nrevezzo said, it may be true.

thanks
Nivas

baabi_26 · Post by **baabi_26** » Tue Jan 25, 2005 3:48 pm

Hi Nivas,

I have tried the same in one of my jobs. If the execution mode is parallel the job will always fail when run on two or more nodes. The workaround is to make it sequential and run the job. I have to agree that the job will be slow. Multiload stage would become bottle-neck.But i am not sure if there is any other way to make it work. Did you try using Enterprise stage?

Thanks
Naveen

DSXchange