7 replies to this topic

[lmotu]

Hello Sirs,

at our company we have to start HDP-sessions which run over two MUs. But the exchange from one MU to the other lasts very long.

So I created a simple test session called "s_test", which contains a uproc, named "test". The code of this uproc is only "exit 0".

The session is build like : START- TEST (MU:SERVER1) - TEST (MU:SERVER2)

At MU1 the session (and uproc START and TEST) starts and ends immediately and correctly.
But the start of this session at MU2 is delayed, up to 40 minutes.

We looked at the universe.log-files, they are quite short. We do a reorgaisation on both Servers.
In the past, we had the same problem and we was briefed by you to razfic the files
u_fecl50 X
u_fecd50 X


We do that again, but without success.
Whats wrong? What can we do to increase the exchange/launch speed to a normel value?

OS=UNIX
uxversion: 5.33



Thank you
lmotu

Edited by men, 23 July 2009 - 09:36 AM.

[men]

Hello,

can you check the network configuration:

Let's imagine that SERVER1 & SERVER2 are the Dollar Universe Node names of both servers. They need to be present in the Node table on both sides as well as the 2 MUs (MU1 & MU2 must be defined in the MU tables on both sides).

On SERVER1, does file uxsrsrv.sck in mgr contain both SERVER1 & SERVER2 ?
On SERVER2, does file uxsrsrv.sck in mgr contain both SERVER1 & SERVER2 ?

In the uxsrsrv.sck, is the correspondance with the Dollar Universe Node name based on the hostname of the servers or their IP address?

If you run a command such as "uxlst SES NODE=SERVER1" from SERVER2 and "uxlst SES NODE=SERVER2" from SERVER1 do you get an immediate or delayed answer?

Has the razfic of u_fecl50 & u_fecd50 taken place on both Nodes, Dollar Universe being stopped?

Last but not least, you may try to clean the exchange & job event files with either the "uxresnod" utility or the "uxclenod" utility. You can get more information on them in the Adminitraor manual (P. 71 & 72).

If none of these actions help, I guess you are left with opening a call at Dollar Access Technical Support...

Michel

[lmotu]

men, on July 23 2009, 12:59 PM, said:

Hello,

can you check the network configuration:

Let's imagine that SERVER1 & SERVER2 are the Dollar Universe Node names of both servers. They need to be present in the Node table on both sides as well as the 2 MUs (MU1 & MU2 must be defined in the MU tables on both sides).

On SERVER1, does file uxsrsrv.sck in mgr contain both SERVER1 & SERVER2 ?
On SERVER2, does file uxsrsrv.sck in mgr contain both SERVER1 & SERVER2 ?

In the uxsrsrv.sck, is the correspondance with the Dollar Universe Node name based on the hostname of the servers or their IP address?

If you run a command such as "uxlst SES NODE=SERVER1" from SERVER2 and "uxlst SES NODE=SERVER2" from SERVER1 do you get an immediate or delayed answer?

Has the razfic of u_fecl50 & u_fecd50 taken place on both Nodes, Dollar Universe being stopped?

Last but not least, you may try to clean the exchange & job event files with either the "uxresnod" utility or the "uxclenod" utility. You can get more information on them in the Adminitraor manual (P. 71 & 72).

If none of these actions help, I guess you are left with opening a call at Dollar Access Technical Support...

Michel

Hello Michel,

all, you descripted was checked and is ok. The exchange/launch of uprocs runs, but extremly delayed.

uxlst returns the answers immediate from both Servers.

I run uxclenod, razfic u_fecl50/u_fecd50 and even uxcleevt - but not uxresnod yet.


Last one is now running (and still running for 15 minutes).
message: "Start purge on files 20090723 15231300"

Thank you
lmotu

[men]

Hello again lmotu,

thanks for the update.

I would recommend you run a reorganization again as soon as you can after having run the utilities. I would favour an offline reorganization over an online one.

Michel

[Ceceil]

Hello lmotu


Did the things you do alleviate your timing issue? I am experiencing the same problems with a session between nodes and was wondering if these steps helped.

Thank you

Ceceil

[lmotu]

men, on July 23 2009, 03:13 PM, said:

Hello again lmotu,

thanks for the update.

I would recommend you run a reorganization again as soon as you can after having run the utilities. I would favour an offline reorganization over an online one.

Michel

Hello michel,

sorry for delayed answering, I was on holiday.
Yes we do a reorganisation and now it runs well!

Thank you for your service.
lmotu

Edited by lmotu, 17 August 2009 - 02:13 PM.

[men]

Good news then

Glad I could help!

Michel

[lmotu]

Ceceil, on August 6 2009, 04:20 PM, said:

Hello lmotu


Did the things you do alleviate your timing issue? I am experiencing the same problems with a session between nodes and was wondering if these steps helped.

Thank you

Ceceil

hello ceceil,

believe it or not! This solved our timing problem.

Best regards
lmotu

Edited by lmotu, 17 August 2009 - 02:16 PM.