Big difference in Transfer of specs from Build machine to I-Series Lpar

[PMatino]

I am having a problem with the transfer of specs from Build machine to I-Series Lpar during a Full Package build.
Below are the timestamps from the Server build log in the transfering of specs. Everything else in the build is the same as it ever was. The deployment server to SCNJ400A the Dev Lpar is fine but from the Deployment server to SCNJ400B(QA) or SCHJC400 (Prod) the transffer rate is ridiculous. Please see below. Any Ideas on what may cause this. The amout of time to transfer is constant though numerous builds at different times which leads me to believe it is not network traffic.


Production 6 1/2 hours

Sat Apr 20 12:00:16 -

Sat Apr 20 12:00:20 - Attaching local repository.

Sat Apr 20 12:00:51 - Local repository SPEC__PD0420FP attached.

Sat Apr 20 12:00:51 - Copying XML specs to datasource Central Objects - PD812.

Sat Apr 20 12:00:51 - Copying XML specs from the build machine to the server spec database datasource Central Objects - PD812.

Sat Apr 20 12:00:51 -

Sat Apr 20 12:00:51 - Creating spec repository in the spec database datasource Central Objects - PD812.

Sat Apr 20 12:01:28 - Spec repository created.

Sat Apr 20 12:01:28 -

Sat Apr 20 12:01:28 - Copy spec table MANIFEST from the build machine to the server spec repository Central Objects - PD812.

Sat Apr 20 12:01:29 - Copying spec table.

Sat Apr 20 12:01:29 - Number of records to copy: 1

Sat Apr 20 12:01:32 - Number of records copied: 1.

Sat Apr 20 12:01:32 - Spec table copied

Sat Apr 20 12:01:32 -

Sat Apr 20 12:01:32 - Copy spec table jdeblc from the build machine to the server spec repository Central Objects - PD812.

Sat Apr 20 12:01:33 - Copying spec table.

Sat Apr 20 12:01:33 - Number of records to copy: 18989

Sat Apr 20 12:23:00 - Number of records copied: 18989.

Sat Apr 20 12:23:00 - Spec table copied

Sat Apr 20 12:23:00 -

Sat Apr 20 12:23:00 - Copy spec table fdaspec from the build machine to the server spec repository Central Objects - PD812.

Sat Apr 20 12:23:00 - Copying spec table.

Sat Apr 20 12:23:00 - Number of records to copy: 439683 THIS took 6 ½ hours to complete. In the end the transfer did not finish the rest of the spec files.


Development 1hour

Fri Mar 15 13:33:57 - Copy spec table MANIFEST from the build machine to the server spec repository Central Objects - DV812.

Fri Mar 15 13:33:57 - Copying spec table.

Fri Mar 15 13:33:57 - Number of records to copy: 1

Fri Mar 15 13:34:00 - Number of records copied: 1.

Fri Mar 15 13:34:00 - Spec table copied

Fri Mar 15 13:34:00 -

Fri Mar 15 13:34:00 - Copy spec table jdeblc from the build machine to the server spec repository Central Objects - DV812.

Fri Mar 15 13:34:01 - Copying spec table.

Fri Mar 15 13:34:01 - Number of records to copy: 19002

Fri Mar 15 13:36:15 - Number of records copied: 19002.

Fri Mar 15 13:36:15 - Spec table copied

Fri Mar 15 13:36:15 -

Fri Mar 15 13:36:15 - Copy spec table fdaspec from the build machine to the server spec repository Central Objects - DV812.

Fri Mar 15 13:36:15 - Copying spec table.

Fri Mar 15 13:36:15 - Number of records to copy: 444433

Fri Mar 15 14:29:49 - Number of records copied: 444433.

Fri Mar 15 14:29:49 - Spec table copied

 

[Tom_Davidson]

I have the same issue with my 8.12 instance. I have been playing with the ODBC settings and so far haven't had a whole lot of success. I have tried changing the LOB threshold, and record blocking with not much success.

Strangely I don't have this issue with my 9.0 install. They are both on the same AS/400, subnet and tools release (8.98.4.4).

If you find a way to speed it up, please let me know.

One question I do have is your deployment server physical or virtual? In my case 8.12 is virtual and 9.0 is physical, but I had the same issue in 8.12 when it was physical.

The only real difference I can think of is that 8.12 uses SQL and 9.0 uses Oracle on the deployment server.

Tom

 

[jdel6654]

Check the TCP buffer sizes for inbound and outbound (CHGTCPA).

For the more current boxes these should be a minimum of 1MB for 1GB network.

 

[jdelisths]

Can you try copying a large file (about 100 Mg) from the deployment server to an IFS folder on both servers to see if there is any difference?