Strange Datastage job behavior!

[splayer]

I have a sequence job which has several job activity stages and execute command stages. Sometimes, when I execute the sequence job, one of the jobs would just hang. No error messages at all. I have to stop the sequence job. When I reset the jobs and restart, the exact same job runs perfectly fine and immediately without any problems. Has anybody seen this?

[DSguru2B]

Keep your logs in check and perfom cleanup of &PH& directory regularly. It might not be stuck and actually running or taking time to startup, but the status update might be taking time.

[abc123]

What is &PH& directory?

[us1aslam1us]

check this link..

viewtopic.php?t=105949&highlight=COMO

sAM

[splayer]

So is &PH& a directory? How do I know its value?

[chulett]

Yes, it is a directory - the PHantom directory, a directory leveraged by the background (i.e. 'phantom') processes each job runs as. There's one in each Project directory. What do you mean by "how do I know its value"?

[splayer]

When I went to the project directory, I saw the following:

D_&COMO&
&COMO&
&SAVEDLISTS&

but I did not see something like &PH&. When I tried to do:
cd &COMO&
-----------------------------------------------------------------------------
I got the following message:
[1] 15295
[2] 15296
myid@MyProject>-bash: cd: cd: No such file or directory
-bash: COMO: command not found

[1]- Exit 1 cd cd
[2]+ Exit 127 COMO
-----------------------------------------------------------------------------

If these are directories and I should clear whatever is in them, shouldn't I be able to go in them?

[ArndW]

splayer - check again, if you don't find a &PH& then perhaps you (or someone else) deleted the directory. Do you have a D_&PH& entry in the project?

You need to execute "cd \&COMO\&' at the command line, as UNIX sees the ampersand as a special character.

[splayer]

ArndW, so should I be clearing all of these folders every time before I run my job sequence?

[michaeld]

I don't have an answer, but I have had this problem before. Lots of times in the windows version. It is usally due to a bug in the job that causes it to use up a lot of memory. However sometime there is no reason for it. I've contacted IBM support about it and they had no answers. Other then clear your log and stuff like that, which was not the problem in my case. I ended up remaking the job (the same way) and it became stable. I think that datastage jobs can become corrupt. I don't know how or why, but it has happed a few times before and usually saving it as another job will fix it.

[splayer]

michaeld, this is not job specific. Exactly the same job works perfectly right next minute. This happens with jobs at random.

[ray.wurlod]

Any kind of "hang" usually means waiting for a resource. Next time this occurs check to see if any kind of lock is held on the job sequence or controlled activity that appears to be hanging. Also check to see that there's plenty of free space in the file systems /tmp and those identified by the UVTEMP configuration parameter and those in your parallel job configuration file as resource disk and scratchdisk (particularly the latter).

[splayer]

So I should do/check for the following:

1) In the &PH& directory, I should delete all DSD.RUN and DSD.StageRun files. I see some DSD.OshMonitor files. Should I delete them as well? Are there any side effects?

2) I go to /tmp folder and delete anything I can in there. I see some files there but they are very small in size.

3) I go to the Scratch folder listed in the configuration file and clear everything in there.

4) In my ResourceDisk directory, I have the binary child files of datasets. Should I go ahead and delete these datasets? I need these datasets further in my process.

5) My UVTEMP directory is defined as /tmp. I have already dealt with it #2 above.

I would appreciate if you could comment on this. Thanks.