SQL1652N : File I/O error occurred

DSguru2B · Post by **DSguru2B** » Tue Jan 30, 2007 9:26 am

Try decreasing the number of parallel events one by one and try to see on what number the error goes away. Just for debugging purposes.

Krazykoolrohit · Post by **Krazykoolrohit** » Tue Jan 30, 2007 12:08 pm

DSguru2B wrote:Try decreasing the number of parallel events one by one and try to see on what number the error goes away. Just for debugging purposes.

Its in production. Lots of processes to be followed even to make one change in the sequencing. Problem is that it only shows issues when ran with production data. It runs perfectly fine with dev data. thats why i think its some performance issue. maybe not with the job but with datastage server itself.

DSguru2B · Post by **DSguru2B** » Tue Jan 30, 2007 12:21 pm

How many times did your try re-running your job? Did it give the same error every single time?

Krazykoolrohit · Post by **Krazykoolrohit** » Tue Jan 30, 2007 12:35 pm

re run of the job makes it run fine. It fails only when it is run with sequencer. IT fails everytime i run it using the sequence.

DSguru2B · Post by **DSguru2B** » Tue Jan 30, 2007 12:40 pm

Are you using any more Bulk Loader jobs in this sequence job which are fired in parallel? I am just guessing here as I am running out of ideas. The Bulk Loder message file, does it contain any helpful messages?

I_Server_Whale · Post by **I_Server_Whale** » Tue Jan 30, 2007 12:40 pm

Is the sequential file that is used in this job being used in any other Job_Activity in your sequence? Can you check if other job activities are using this sequential file?

Whale.

narasimha · Post by **narasimha** » Tue Jan 30, 2007 12:44 pm

Introduce some delay/sleep between the jobs in your sequence, so that the files get exclusive access.

Krazykoolrohit · Post by **Krazykoolrohit** » Tue Jan 30, 2007 12:45 pm

DSguru2B wrote:Are you using any more Bulk Loader jobs in this sequence job which are fired in parallel? I am just guessing here as I am running out of ideas. The Bulk Loder message file, does it contain any helpful messages?

I have already ran out of ideas. The bulk loader message file has not much information as the load fails before even 1 row is loaded.

There are around 12 bulk loader jobs which run in parallel but with different tables as targets.

Krazykoolrohit · Post by **Krazykoolrohit** » Tue Jan 30, 2007 12:46 pm

I_Server_Whale wrote:Is the sequential file that is used in this job being used in any other Job_Activity in your sequence? Can you check if other job activities are using this sequential file?

Whale.

No the sequential file is not used by any other job/activity. the folder in which the file is being accessed by many jobs though

Krazykoolrohit · Post by **Krazykoolrohit** » Tue Jan 30, 2007 12:47 pm

narasimha wrote:Introduce some delay/sleep between the jobs in your sequence, so that the files get exclusive access.

exclusive access to what? how will inducing sleep between jobs help?

narasimha · Post by **narasimha** » Tue Jan 30, 2007 12:54 pm

Rohit,
You could write a routine to sleep for x seconds, call it in between your jobs in the Job sequence.
When I said exclusive access, I mean wait for the resources involved to get freed up.

DSguru2B · Post by **DSguru2B** » Tue Jan 30, 2007 1:11 pm

Krazykoolrohit wrote:
There are around 12 bulk loader jobs which run in parallel but with different tables as targets.

Are you providing different TEMP_DIR's for all these jobs in the DB2 Load Stage? They are all trying to create the same set of command files. If their folder is the same then thats your problem right there. Also make sure that the message files have unique names.

Krazykoolrohit · Post by **Krazykoolrohit** » Wed Jan 31, 2007 11:15 am

DSguru2B wrote:
Krazykoolrohit wrote:
There are around 12 bulk loader jobs which run in parallel but with different tables as targets.
Are you providing different TEMP_DIR's for all these jobs in the DB2 Load Stage? They are all trying to create the same set of command files. If their folder is the same then thats your problem right there. Also make sure that the message files have unique names.

I think you have a point here. This may be the reason. for some strange reason our support team wants all jobs to write to the same log file. will correct it and see how it goes. thanx

DSguru2B · Post by **DSguru2B** » Wed Jan 31, 2007 12:43 pm

Krazykoolrohit wrote: for some strange reason our support team wants all jobs to write to the same log file.

Possible, but it has to be sequential. For parallel processing, they have to be seperate files. Perhaps your support team needs to know that simultaneous writes to the same file is not possible, whether its the command files or log files.

Krazykoolrohit · Post by **Krazykoolrohit** » Wed Jan 31, 2007 1:52 pm

DSguru2B wrote:
Krazykoolrohit wrote: for some strange reason our support team wants all jobs to write to the same log file.
Possible, but it has to be sequential. For parallel processing, they have to be seperate files. Perhaps your support team needs to know that simultaneous writes to the same file is not possible, whether its the command files or log files.

They have their wierd ways of forcing things which doesnt make any sense. its a long story. I will keep this post updated. it may take few days to test the changes done as the jobs are in prodn