time out

hne · Post by **hne** » Wed Nov 21, 2007 2:11 am

Is it possible to increase the time out so that datastage tries longer than 60 sec to start a job.

ray.wurlod · Post by **ray.wurlod** » Wed Nov 21, 2007 5:02 am

There ought to be no need to.

That way lies madness.

You'll want to increase it towards infinity, then you'll complain that your jobs don't reach a running state.

Reduce the load on your machine(s) so that jobs can start within 60 seconds.

hne · Post by **hne** » Wed Nov 21, 2007 8:40 am

I heard the rumour that the use of basic transformers can overload the server. Is this correct?
I ask this because other ds projects than mine runs on the same server. Even when the cpu and memory usage of the server is very low, sometimes we have constant time out errors. The other projects uses a lot of basic transformers in parallel jobs.

hne · Post by **hne** » Wed Nov 21, 2007 8:45 am

I heard the rumour that the use of basic transformers can overload the server. Is this correct?
I ask this because other ds projects than mine runs on the same server. Even when the cpu and memory usage of the server is very low, sometimes we have constant time out errors. The other projects uses a lot of basic transformers in parallel jobs.

ray.wurlod · Post by **ray.wurlod** » Wed Nov 21, 2007 2:08 pm

Where did you hear that rumour?

It's not particularly true except to the same extent as most other stage types - too many of anything will overload the machine. Too many jobs executing at the same time will overload the machine.

hne · Post by **hne** » Thu Nov 22, 2007 6:38 am

The rumour was internal but not proved, but how can you explain time outs if prstat -a gives the following :

PID USERNAME SIZE RSS STATE PRI NICE TIME CPU PROCESS/NLWP
17282 PPdtsusr 28M 14M sleep 59 0 0:00.11 0.1% dsapi_slave/1
11750 root 214M 34M sleep 59 0 1:51.15 0.1% java/31
25618 PPdtsusr 26M 12M sleep 22 0 0:01.42 0.1% dsapi_slave/1
17467 PPdtsusr 27M 12M sleep 59 0 0:00.12 0.0% dsapi_slave/1
21341 root 3272K 2720K sleep 3 0 0:00.00 0.0% se.sparcv9.5.8/1
24185 PPdtsusr 27M 12M sleep 59 0 0:00.14 0.0% dsapi_slave/1
21344 PPdtsusr 1368K 1072K cpu34 2 0 0:00.00 0.0% prstat/1
1196 root 4264K 3912K sleep 1 0 0:18.32 0.0% nscd/13
21339 root 1904K 1408K sleep 1 0 0:00.00 0.0% Solaris_RICHPse/1
24398 PPdtsusr 26M 12M sleep 54 0 0:20.11 0.0% dsapi_slave/1
1689 root 37M 14M sleep 2 0 0:24.36 0.0% java/17
17281 PPdtsusr 6864K 3072K sleep 59 0 0:00.00 0.0% dsapi_server/1
11756 root 207M 32M sleep 59 0 1:22.42 0.0% java/27
3281 root 15M 5384K sleep 29 0 0:07.30 0.0% opcmona/7
1068 root 5480K 3496K sleep 3 0 1:00.11 0.0% automountd/5
NPROC USERNAME SIZE RSS MEMORY TIME CPU
66 PPdtsusr 903M 430M 5.4% 0:51.57 0.2%
96 root 1250M 356M 4.5% 7:19.02 0.1%
1 daemon 2656K 1968K 0.0% 0:00.00 0.0%
Total: 163 processes, 584 lwps, load averages: 0.07, 0.50, 0.90

ray.wurlod · Post by **ray.wurlod** » Thu Nov 22, 2007 2:40 pm

Waiting for some other resource, such as I/O, network, a semaphore or database lock, will not consume resources (and therefore will not contribute to prstat and similar reports). But it might still engender a timeout. For example, trying to start a section leader on another machine on a saturated LAN can take longer than is permitted.

hne · Post by **hne** » Fri Nov 23, 2007 3:58 am

The most intresting part of your answer is not readable.

ray.wurlod · Post by **ray.wurlod** » Fri Nov 23, 2007 4:34 am

Yes it is. It is written in very clear English.

For less than 30c per day (currently about Rs12) you can purchase premium membership which would permit you to read complete posts by the five premium posters. Revenue from premium memberships is totally dedicated to meeting the hosting and bandwidth costs of DSXchange - no-one, including the premium posters - is paid for posting.