Page 1 of 1
Multi instance Sequence job failing
Posted: Mon Sep 22, 2008 10:35 am
by swades
Hi All,
We have a multi instance sequence job. We are calling 12 instances at a same time. Every time few of instance (not the same every time) instance failing with following fatal error.
Calcom_Daily_Dataload_Master_Seq.GMO.JobControl (DSSendMail): Error when executing command: /texsrvr01/Ascential/DataStage/Projects/GTS_GLOBAL/sendmail_Calcom_Daily_Dataload_Master_Seq.sh
*** Output from command was: ***
SH: /texsrvr01/Ascential/DataStage/Projects/GTS_GLOBAL/sendmail_Calcom_Daily_Dataload_Master_Seq.sh: not found
If we run instance one by one, all the instance running fine. But if we run all the instances same time, the sequence job is failing at notification activity for sending success mail(rest of job running fine).
I did search the forum. But could not found any reasonable clue.
Thanks in advance.
Posted: Mon Sep 22, 2008 10:46 am
by chulett
That sounds like a bug that should be reported to your official support provider.
Each Notification Activity creates a shell script that it executes and then deletes. Sounds like all of the instances are using the same filename and depending on the timing, another instance deletes the script that some other instance is trying to run.
Posted: Mon Sep 22, 2008 11:33 am
by chulett
I would still report it as a bug if you haven't done so already.
Any way to consolidate the notifications? Have one master sequence that handles all other invocations and the emails, perhaps? You could also write your own email routine, invoke 'sendmail' or whatever you like with DSExecute(), it shouldn't have any kind of multiple instance issues that I can think of.
Posted: Mon Sep 22, 2008 11:56 am
by swades
Is there anything - I can change at project level and things work fine. What I am looking for is - take the control from DS and on the fly creates this shell script with invocation id in script name. So it will not conflict with other instance.We have few jobs already moved to production and facing this issue. So we would like to do minimum level of code change.
Posted: Mon Sep 22, 2008 11:57 am
by swades
Is there anything - I can change at project level and things work fine. What I am looking for is - take the control from DS and on the fly creates this shell script with invocation id in script name. So it will not conflict with other instance.We have few jobs already moved to production and facing this issue. So we would like to do minimum level of code change.
Posted: Mon Sep 22, 2008 12:06 pm
by chulett
No need to take that same scripted approach that the stage does. Look into the Routine Activity idea that I posted. Replace the Notification Activity stage with a Routine Activity stage that does the emailing. Should be simple enough to setup and won't have the multi-instance problem that you are seeing now.
You can use DSSendMail or execute your email client of choice (like sendmail) directly that way.
Posted: Mon Sep 22, 2008 3:31 pm
by swades
Thanks Craig-
I have created below routine. This works fine for me, but still the question is if this routine called from routine activity in multi instance sequence job, might be two routine instances run at same time. Is this can be a problem?
Code: Select all
Arg="From:":SenderMailId:"\nTo:":ReceptionMailId:"\nSubject:":JobStatus:": ":ProjectName:"-->":JobName:".":ErrorSource:" (":HostName:")\nBody:":BodyMessage
CALL DSU.DSSendMail(Arg,ErrorCode)
Ans = 0
Posted: Mon Sep 22, 2008 3:52 pm
by chulett
I don't
believe so, but if it does I'm sure you'll let us know.
If you have that problem, try switching it to use DSExecute() and call 'sendmail' directly, that should work fine.
Posted: Mon Sep 22, 2008 5:45 pm
by swades
chulett wrote:I don't
believe so, but if it does I'm sure you'll let us know.
no room for experiments.
Thanks very much Craig for your time and support.
Posted: Mon Sep 22, 2008 5:47 pm
by swades
chulett wrote:I don't
believe so, but if it does I'm sure you'll let us know.
no room for experiments.
Thanks very much Craig for your time and support.
Posted: Mon Sep 22, 2008 5:49 pm
by swades
chulett wrote:I would still report it as a bug if you haven't done so already.
Need to check with Admin team, if they didn't , I will report to IBM.