9.1 slow connection times

PeteM2 · Post by **PeteM2** » Tue Oct 22, 2013 5:17 am

We have migrated from 7.5.2 to 9.1 and have noticed when running jobs from the command line using the DSJOB command to find out the previous status of a job, we are seeing a 3 to 5 minute delay compared to 7.5.2.

I was wondering whether this was due to the different architecture in 9.1 and the fact that prior to connecting to the 'engine' authentication is performed on the service tier first. Whereas on 7.5.2 the connection id directly to the engine.

any ideas on the delay?

priyadarshikunal · Post by **priyadarshikunal** » Tue Oct 22, 2013 5:45 am

There may be a little delay but I don't think it will be 3 to 5 minutes anyways. The start-up timeout is much less than 3 minutes unless overridden. Did you mention the delay per dsjob command or for a script with multiple dsjob operations?

How loaded is your server? Also version 8.0 onwards datastage installation needs more capacity than 7.5.2.

Also try using DOMAIN=NONE and pass engine tier user details (unix user details) instead of service tier user details and see if it reduces the delay time.

ArndW · Post by **ArndW** » Tue Oct 22, 2013 6:04 am

Are you using the operations console and workload management at 9.1? That might cause delays between jobs are started and when they actually get executed (in the interim they have the status of "queued"). I've found that my clients seem to be a bit sluggish at 9.1 compared to 8.7, but that execution times are roughly the same.

PeteM2 · Post by **PeteM2** » Tue Oct 22, 2013 8:21 am

The delay is for one DSJOB operation that is obtaining the job status for the previous run.

We are using the ops console workload management feature but do not see anything queueing up.

priyadarshikunal · Post by **priyadarshikunal** » Tue Oct 22, 2013 8:24 am

I forgot about WLM in 9.1 entirely

ArndW · Post by **ArndW** » Tue Oct 22, 2013 9:59 am

What about the max number of jobs running setting, I am not sure tasks shows as "queued" if that limit is reached, they might just wait until they get a slot.

What is your dsjob.exe syntax?

PeteM2 · Post by **PeteM2** » Wed Oct 23, 2013 5:19 am

below is the dsjob command

$DSBin/dsjob -authfile $scriptdir/.DSAuthFile.ini -jobinfo PROD jobname

PeteM2 · Post by **PeteM2** » Wed Oct 23, 2013 5:21 am

below is the dsjob command

$DSBin/dsjob -authfile $scriptdir/.DSAuthFile.ini -jobinfo PROD jobname

ArndW · Post by **ArndW** » Wed Oct 23, 2013 6:20 am

I am at a loss to explain why it takes so long to return that information. Are all your tiers on the same physical machine and how busy is it when you execute the command. If on separate machines, could one of them be at 100% CPU or I/O and thus blocking the command? I've never noticed any delays in just getting the jobinfo (although I've never used the command-line authentication, but I don't quite see how that could be making a difference here).

PeteM2 · Post by **PeteM2** » Wed Oct 23, 2013 7:25 am

All the tiers are on the same physical machine

jdinero · Post by **jdinero** » Thu Oct 31, 2013 10:32 am

I found this on the IBM set, might be related:

http://www-01.ibm.com/support/docview.w ... wg1JR47809

ArndW · Post by **ArndW** » Thu Oct 31, 2013 11:37 am

Nice find - that might be the cause, we'll see if the OP returns and can confirm their port usage.