On a frequent basis my jobs get locked at the Server

pravesh_kushwaha · Post by **pravesh_kushwaha** » Fri Sep 10, 2004 12:43 am

We have just installed Ascential DataStage 6.01 on Hp Ux server.

For all the projects we are facing the following problem.

On a frequent basis our jobs are getting locked on the DataStage Server.
In the administrator we had set the Inactivity Timeout setting to do not timeout but still our jobs are getting locked.

Can someone tell us why this is taking place and how to resolve it.

denzilsyb · Post by **denzilsyb** » Fri Sep 10, 2004 1:02 am

pravesh_kushwaha wrote: For all the projects we are facing the following problem.

How stable is the network? It could be that developers have jobs open, network conn is lost briefly and on UNIX the job is still 'open' but the client has lost the handle on the job.

To test this, have a client ping -t server while developing on DS, and as soon as a job is 'lost' check the output of the ping. This would verify if the network is the culprit.

ray.wurlod · Post by **ray.wurlod** » Fri Sep 10, 2004 1:10 am

"Locked by another process" can be when you have the same job design open in another window, or another user has it open. It can also occur if the job is being monitored or, in some cases, executed.

"Locked by another process" can also occur if the process owning the lock is made defunct, for example the PC is shut down or disconnected, or the process on the server is killed.

The recovery is fairly straightforward; best is to make it automatic by enabling the DataStage deadlock daemon, which wakes periodically to check for deadlocks and also to clean up after defunct processes.

There is also a manual strategy by which an administrator can UNLOCK locks held by defunct processes.

Searching this forum for those topics should see you well on your way.

pravesh_kushwaha · Post by **pravesh_kushwaha** » Fri Sep 10, 2004 1:19 am

Can you explain how to enable the DataStage DeadLock Daemon.

Can you also give more information regarding the manual strategy for unlocking the job.

Regarding the stability of network : Network is highly stable and we are having a dedicated unix server for DataStage. Our job is getting locked even if when it was opened previuosly proper closing procedure was followed to exit from the job.

denzilsyb · Post by **denzilsyb** » Fri Sep 10, 2004 1:24 am

ray.wurlod wrote:best is to make it automatic by enabling the DataStage deadlock daemon, which wakes periodically to check for deadlocks and also to clean up after defunct processes

Ray - this sounds great. which manual do we read to get on this!

denzilsyb · Post by **denzilsyb** » Fri Sep 10, 2004 5:09 am

pravesh_kushwaha wrote: Can you also give more information regarding the manual strategy for unlocking the job.

(this is how it is for Server Jobs. Paralell might be different)

Using DS administrator.
select project wherein the job is locked
select COMMAND button
type DS.TOOLS
select 5 (Administrative tasks)
select 4 (list all locks)
you might need to resize the window to view all the data
write down the PID of the job that is locked
once all have been listed you will return to the main menu

select 7 (clear locks held by process)
type the PID of the job that is locked
confirm with 'Y'

and your job should be cleared.

by the way - did you restart the DS services after setting the 'do not timeout' feature?

pravesh_kushwaha · Post by **pravesh_kushwaha** » Fri Sep 10, 2004 7:45 am

Hi Denzilsyb,
Thanks for providing that info, but the DS.TOOLS utility is not listing neither any processes nor any locks if we use the options available.

The options are not working.

You have gone through the mine other post at DSXCHANGE in which i have said that clean up resources is not working.

The DataStage server is not able to retrieve any information about the process_id or locks occupied by any Job.

I got another way of unlocking the job but that to is giving error:

Another Method:

On Unix Prompt DataStage Directory

./dsenv
cd bin
dssh
LOGTO UV
UNLOCK INODE inode_num USER user_num ALL

In the above command ALL specifies releasing all locks

I am getting error insufficient privilege for unlocking inode.

denzilsyb · Post by **denzilsyb** » Fri Sep 10, 2004 8:11 am

pravesh_kushwaha wrote: I am getting error insufficient privilege for unlocking inode.

does your user have permisions to do this task?

Im not sure you want to release all locks - there might actually be valid locks out there. Have you looked at the user reference manual for BASIC. I believe the site is documented on the forum somewhere, but if you want i could email you the manual - just send me a message.

The method you describe above is the first I have seen - thanks for that.

just a question again - did you restart the services when you checked 'do not time out'?

pravesh_kushwaha · Post by **pravesh_kushwaha** » Fri Sep 10, 2004 9:40 am

How can we check whether user is having the required permissions to do these tasks of unlocking or killing processes.

Can anyone tell more about these privileges.

ray.wurlod · Post by **ray.wurlod** » Fri Sep 10, 2004 5:05 pm

To enable automatic starting of the DataStage deadlock daemon go to your DataStage Engine directory on the server and find the file called dsdlockd.config. Edit this file, changing start=0 to start=1. It will start next time DataStage is started.

To start the deadlock daemon manually you need to be logged in with superuser privilege. Get into the DataStage environment in the DataStage Engine directory and issue the command DEADLOCK.MENU. Option 2 starts the deadlock daemon, first prompting you for the sleep interval.

Deadlock daemon keeps a log file called dsdlockd.log in the DataStage Engine directory (the same directory where dsdlockd.config is).

ray.wurlod · Post by **ray.wurlod** » Fri Sep 10, 2004 5:12 pm

denzilsyb wrote:Ray - this sounds great. which manual do we read to get on this!

It doesn't appear in any of the DataStage manuals. You will find it in the IBM manual Administering UniVerse, as the final section of Chapter 9. This manual can be downloaded at no charge from IBM's UniVerse manuals website.

ray.wurlod · Post by **ray.wurlod** » Fri Sep 10, 2004 5:17 pm

pravesh_kushwaha wrote:How can we check whether user is having the required permissions to do these tasks of unlocking or killing processes. Can anyone tell more about these privileges.

UNLOCK is one of the commands reserved for the DataStage Administrator. Who this is depends on how you installed DataStage; in particular whether you have chosen to use impersonation mode.

In your case I would guess that you need to be logged in either as dsadm or as root.

On Windows systems the DataStage administrator is any member of the Administrators group.

Read page 1-1 of the manual Administrator Guide (dsadmgde.pdf), particularly the section entitled "Who Can Administer DataStage?"

ray.wurlod · Post by **ray.wurlod** » Fri Sep 10, 2004 5:19 pm

Please note that I have made three separate posts immediately above. Don't read just the last one!

pravesh_kushwaha · Post by **pravesh_kushwaha** » Sun Sep 12, 2004 4:09 am

Thanks ray.world.

When i executed command DEADLOCK.MENU it gives error command not found.

I have tried all the commands by logging as root as well as dsadm.

chulett · Post by **chulett** » Sun Sep 12, 2004 8:11 am

You need to 'LOGTO UV' before the command will work.

DSXchange

On a frequent basis my jobs get locked at the Server

On a frequent basis my jobs get locked at the Server

Re: On a frequent basis my jobs get locked at the Server

DataStage Deadlock Daemon