Problem with accessing sequential file in a shared container

Jdrost · Post by **Jdrost** » Tue Sep 12, 2006 7:24 am

Hi,

I have developed a shared container to add error-messages to a error-log (sequential-file). The name of the file is hard-coded. I use this container in several (upto 20) server-jobs. When I run the sequence with all these server-jobs running parallel at a certain moment the job fails with the following message:

'Inflow_DB_DBSTP_clean..AddErrorLog.ERROR_LOG.ERROR_LOG_OUT: DSD.SEQOpen Unable to create file C:\LPGP_D_CRISKM_PROJECT\ErrorLog.csv.'

I think this error is a result of a job trying to access the error-file when it is locked for update by another job (is it?). The only solution to solve this problem I can think of is to execute the jobs in sequence instead of in parallel.

Does anyone know a better solution for this problem?

kcbland · Post by **kcbland** » Tue Sep 12, 2006 7:25 am

Two or more processes cannot WRITE to a sequential file simultaneously. Consider using a parameter in the file name.

DeepakCorning · Post by **DeepakCorning** » Tue Sep 12, 2006 7:30 am

If you are simultaneously triggering two jobs with the shared container in it then it will give the same error as they can write to the seq file at the same time. Try using Hashed File.

Jdrost · Post by **Jdrost** » Tue Sep 12, 2006 7:49 am

DeepakCorning wrote:If you are simultaneously triggering two jobs with the shared container in it then it will give the same error as they can write to the seq file at the same time. Try using Hashed File.

Why use a hash-file? Is it possible to read and write simultaneously from multiple jobs to a hash-file?

kcbland · Post by **kcbland** » Tue Sep 12, 2006 8:02 am

Hashed files can handle multiple processes reading/writing, as long as they are not accessing the same row. Avoid using the locking mechanisms as it is a bad technique.

DeepakCorning · Post by **DeepakCorning** » Tue Sep 12, 2006 8:22 am

Yes , Hashed File can be written by many jobs and hence are a good replacement for seq files in a shared container.

ray.wurlod · Post by **ray.wurlod** » Tue Sep 12, 2006 4:28 pm

A Hashed File is simply another mechanism for implementing a database table. So imagine that you're inserting new rows into a database table.

Create the hashed file with a UV stage but edit the DDL so that you get auto-generated key values.

Code: Select all

CREATE TABLE ErrorTable (
   SurrKey INTEGER NOT NULL DEFAULT NEXT AVAILABLE,
   ErrCode VARCHAR(20) NOT NULL DEFAULT '',
   ErrText VARCHAR(254) NOT NULL DEFAULT '',
   JobName VARCHAR(254) NOT NULL DEFAULT '',
   JobStartTimeStamp CHAR(19)
);