Configuration File

[parameswar]

Hi ,

To test maximum loads in DataStage server we executed 8 sequences in parallel where total number of jobs is around 70. Here all the jobs failed with the below errors:
StgRejExpCodes,2: kgefec: fatal error 0
OCI-21503: program terminated by fatal error
OCI-04030: out of process memory when trying to allocate 16408 bytes (top-most SGA,library cache)

StgRejExpCodes,1: /usr/lib/hpux64/dld.so: Mmap failed for the library</dspx00/Ascential/DataStage/DSEngine/java/jre/lib/IA64W/hotspot/libjvm.so> : Not enough space.

StgRejExpCodes,2: Errors in file :
OCI-21503: program terminated by fatal error
OCI-04030: out of process memory when trying to allocate 16408 bytes (top-most SGA,library cache)
----- Call Stack Trace -----
calling call entry argument values in hex
location type point (? means dubious value)
-------------------- -------- -------------------- ----------------------------
----- Argument/Register Address Dump -----
----- End of Call Stack Trace -----

StgRejExpCodes,1: /usr/lib/hpux64/dld.so: Mmap failed for the library</u00/app/oracle/product/client/lib/libclntsh.so.10.1> : Not enough space.


Sine it is 4 CPU, onfiguration file contains 4 Node, each node is pointing 4 file systems in round robin.It is typically as per configuration guide.

I am unable to find where exactly the problem is, Is it OS level or in configuration file or there is a limit in datastage in terms of number of jobs. waiting for your valuavle inputs.


Thanks in advance,
Parameswar

[ArndW]

You would seem to have run out of total virtual memory on your machine but starting that many datastage jobs and processes. If you increase your swap space you will probably get rid of these errors but chances are quite good that your performance will be abyssmal.

You can look up "out of process memory" and "mmap failed"

[parameswar]

You would seem to have run out of total virtual memory on your machine but starting that many datastage jobs and processes. If you increase your swap space you will probably get rid of these errors but chances are quite good that your performance will be abyssmal.

You can look up "out of process memory" and "mmap failed"

Thanks for your response. Is there any thing I need to change configuration file. My configuration file is as follow:

{
node "node1_hpus51"
{
fastname "hpus51"
pools ""
resource disk "/dsxeone/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxetwo/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxethree/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxefour/Ascential/DataStage/Datasets" {pools ""}
resource scratchdisk "/dsxeone/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxetwo/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxethree/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxefour/Ascential/DataStage/Scratch" {pools ""}

}
node "node2_hpus51"
{
fastname "hpus51"
pools ""
resource disk "/dsxetwo/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxethree/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxefour/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxeone/Ascential/DataStage/Datasets" {pools ""}
resource scratchdisk "/dsxetwo/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxethree/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxefour/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxeone/Ascential/DataStage/Scratch" {pools ""}
}
node "node3_hpus51"
{
fastname "hpus51"
pools ""
resource disk "/dsxethree/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxefour/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxeone/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxetwo/Ascential/DataStage/Datasets" {pools ""}
resource scratchdisk "/dsxethree/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxefour/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxeone/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxetwo/Ascential/DataStage/Scratch" {pools ""}
}
node "node4_hpus51"
{
fastname "hpus51"
pools ""
resource disk "/dsxefour/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxeone/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxetwo/Ascential/DataStage/Datasets" {pools ""}
resource disk "/dsxethree/Ascential/DataStage/Datasets" {pools ""}
resource scratchdisk "/dsxefour/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxeone/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxetwo/Ascential/DataStage/Scratch" {pools ""}
resource scratchdisk "/dsxethree/Ascential/DataStage/Scratch" {pools ""}
}
}

Is there any other clue to increase the performance..

Thanks,
Parameswar

[chulett]

Get a bigger boat?

Since it is 4 CPU, configuration file contains 4 Node

CPU != Node.

[ArndW]

The number of nodes does not have anything directly to do with physical CPUs on the machine! If you don't increase your virtual memory on the machine, then change your configuration file to run with less nodes, that has a very direct impact on the number of processes running and thus on the amount of memory required. Try a 3 node configuration. Also, look at the APT_DUMP_SCORE of your jobs to find out how many processes are invoked and see if you can tune your jobs to make better use of resources.