Improving event server 1

[kid123]

Hi,

We are using Mercator 6.7. One event server is dedicated for
audit mechanisms,in this only one realtime inbound map is
also running. Suddenly I could feel that this server
execution becomes slow.

Can anybody generalise,what are all I can look for?

sometimes,when the load increased,even the server gets crashed.

Eventhough I didnot give sufficient information,please give me what is that can be improved upon.

Will provide all the details required as you ask for.

This is actually very high priority task. Any help would be
appreciated.

Thanks.

 

[rarunprasad]

Hi
Event Server can crash due to various reasons. Normally we are asked to send traces and debug information to Mercator for resolution.

Couple of reasons we had problems with the Event SErver crashing were FTP connections taking a long time without timing out, large files being processed and Database locks.

HTH

Arun

 

[kid123]

Hi,

Thanks for the response.

Currently we are using Mercator 6.7.1. Did you ever face
Event server crash due to the dispatch of more Idocs?
B'cos in our scenario, if we dispatch more than 2500,
it just crashes the server.

If you have faced it already,can you please tell us
whats the solution for this?

Thanks in advance.

 

[janhes]

Hi,
In response to your first question, I have experianced a similar problem on Windows 2000.
It was resolved by amending mercatio.ini settings.
We have an audit map which processes around 200,000 logs per day. We found that if the pending queue got over around 100 entries everything ground to a halt.
We overcame the problem using the folowing settings

; Miscellaneous Controls
;
; maximum concurrent maps (-mx). This value needs to be set carefully: if too low
; then CPU may not be used to full potential. If too high, then excessive memory
; use or excessive context switching can reduce overall performance. 20 is a
; reasonable value for an average server. The higher the number of CPUs on the
; server, the greater the number of maps that can be executed concurrently.
MaxThreads=20

PauseTimeout=15000 ; not used in this version

; file time granularity in seconds (-tx)
TriggerTime=1

; launcher log types - e = error, w = warning, s = startup c = cfg summary
LauncherLog=ewsc=WORKFLOWlauncher.txt

;use the following to pause the listener threads if too many init pending occur
InitPendingHigh=30

;use the following to resume the listener threads if paused from InitPendingHigh
InitPendingLow=10


IdleMQS=1
;SLimMQS=0
;HLimMQS=0
;KeepMQS=0
;MinMQS=0

IdleHTTP=10
;SLimHTTP=0
;HLimHTTP=0
KeepHTTP=0
;MinHTTP=0


The IdleMQS was recommended by Ascential support as a setting in an adapter we do not use.
The http adapter is used for sending an email if there is an error in a log file.
I also found that if any of the debug flags are set that I encountered a lot of GPFs

Jan

 

[BocaBurger]

IdleHTTP is for sending e-mails?

First, what platform are you running on? Have you tuned page size & count (this is critical in 6.7 and up).

What about other loads on the server?

When it "crashed" what error was in the logs?

Do you have watchmaxthreads set?

Too many variables to find out the source. Contact Ascential Support.



BocaBurger
<===========================||////////////////|0
The pen is mightier than the sword, but the sword hurts more!