Java server disconnects from Message server - large XI message

[deidre_logan]

hello there,

We have been having fun with our SAP XI production installation. Today we had an interface start at 10:12 on java server1. It seemed like the java server was not running anymore but it is still in a running state in JCMON and in the System Info webpage.

We came to find out the the log /usr/sap/<SID>/SCS<##>/work/dev_ms had the following error message...

[Thr 1] Fri Jul 14 10:29:23 2006

[Thr 1] *** ERROR => MsSCheckConnections: no response from client C3 (J2EE76452

6051) after 300 secs received [msxxserv_mt. 2923]

[Thr 1] *** ERROR => MsSCheckConnections: disconnect now [163.243.49.225] [msxx

serv_mt. 2930]

[Thr 1] MsJ2EE_AddDisconnectedNode: add node [764526051] into disconnect list

[Thr 1] MsJ2EE_CleanDisconnectedNodes: clean disconnect list

The message that was process was an SAP R3(rfc) --> XI --> (file) ... the file we are processing is only 73mb.

The server1 is up and running but I am not sure it is registerd to the Message server.... How can we see if Server1 is been reconnected to the SCS message server?

Thanks for your assistance.

[deidre_logan]

Check the Message Server - OSS note 650729

You can use the msclients test program to check the message server. The program lists all application servers that are logged on to the message server at the time..

msclients name=<SID>

msclients pf=<instance profile>

msclients pf= <instance profile> -d J2EE

Monitor the Message Server - OSS note 636938

You can monitor the message server by using the msprot test program. The program provides various options that are described in this note.

msprot -a -l name=<SID>

msprot -l pf=<instance profile> -d J2EE

Test Logon Groups - OSS note 64015

To test the message server connection, you should use the above-mentioned SAP configuration parameters

lgtst name=SID

using trcfile: dev_lg

list of reachable application servers

-

---

[host2_SID_78] [host2] [13.23.48.21] [sapdpxx] [32xx] [DIA UPD BTC SPO UP2 ICM ]

[host_SID_xx] [host] [13.23.49.25] [sapdpxx] [32xx] [DIA UPD ENQ BTC SPO UP2 ICM ]

list of selectable logpn groups with favorites

-

---

[PUBLIC] [11.11.11.11] [32xx] [640]

[SPACE] [11.24.18.21] [32xx] [640]

OS Data Collection:

HP suggested additional data collection suggestions (run on C/I, 14_32 is batch, 14_33 is SCM)

HP-UX kernel trace. Requires root to run, and needs ~1GB of free space. Collects a 20 sec hp-ux kernel trace. Can be run multiple times.

tusc of process experiencing the problem (we'll have to pick ones we belive to be relevant: enq, msg server, a few dialog). This is a trace of the system calls that a particular process is making.

caliper cstack - similar to tusc, but in this case instead of kernel tracing, this is process tracing. The cstack in particular shows a stack trace for the particular process. Again, running this for one of the work processes as well as the enqueue process (or any other process you think is involved) might be helpful to SAP developers.

Contact DBA on call (if not scripted yet) to run a statspack report for 15-30 minute period. This is an oracle statistics report. Although this is not an HP-UX report, it is often helpful to have this in addition to the kitrace.

[STALANKI]

There are specific settings to be done while processing large files...!...FAQ for file adapter in service market place is the place probably where u can look ...else wise

if u r using large files use abap mapping that z cost effective...

[deidre_logan]

We have the Java heap at 2048 for each of the 3 java servers. I am not sure how much bigger to go in order to process a 500mb file. We have a 24GB HP-UX box. I am thinking of uping the heap to 3072 and see if it helps.