---- Original Message -----
From: "Stricklin, Raymond J" <[EMAIL PROTECTED]>
To: <IBMVM@LISTSERV.UARK.EDU>
Sent: Tuesday, June 03, 2008 11:12 PM
Subject: FCXPER315A
We recently began, under unknown circumstances, receiving this exception
from PTK on one VM LPAR (z800, V5R3):
FCXPER315A Cl1 time slice 1.444 exceeds limit 1.000 (Q1=21 Qx=24)
We received it (with slightly different numbers) once every ten minutes
for about a month, when it stopped under circumstances equally unknown
as those surrounding its beginning. Only one LPAR of the four on the
machine reports this exception.
I don't really know what it's referring to (apart from the dispatcher,
generally), or what its significance might be. Is it of any
significance? Where might I begin looking to determine the ultimate
cause?
Selecting context help for the '85% elapsed time' field on the CPU screen
explains meaning and use of the class 1 elapsed time slice:
'Elapsed time slice which determines the maximum time a
VMDBK may remain in the dispatch list before it is
dropped. This value is continuously adapted for class 1
transactions so that 85% of them can complete within a
single class 1 elapsed time slice, i.e. the value
reflects both the system responsiveness to user demand
and workload characteristics. You should bear this in
mind when comparing the values from different systems. ...'
The messages you saw were probably generated by the initial
threshold setting provided with the sample FCONX $PROFILE:
a value exceeding 1 second would have meant bad response
times on a system with mainly interactive CMS (e.g. PROFS)
users.
However, as the text says, the value depends not only on system
responsiveness but also on your workload, and the sample
threshold may be far off the target for your system.
You should check the actual values for the class 1 elapsed time
slice on your system (see C1ES values on REDISP screen) and
relate them to good/bad performance felt by your users. Then
adapt the threshold value in the 'FC LIMIT C1ES ..' command
in the FCONX $PROFILE to a value where experience has
shown that system performance is getting worse.
Or simply disable threshold checking for C1ES if you cannot
determine a good threshold value ..
Eginhard
