Use the following examples to obtain the appropriate diagnostic
information for system logger. The amount of information requested
in the dumps may be very large. You may need to set your MAXSPACE
on the CHNGDUMP setting to 999 mb before obtaining the logger dumps.
CD SET,SDUMP,MAXSPACE=999M
For structure dumps, verify that the coupling facility has dump
space defined by issuing the following command:
D CF,CFNAME=xxxx
Note: There
are several sample Logger dump parmlib members that can be used as
models for automating the procedures listed below. The samples parmlib
members are shipped in 'SYS1.SAMPLIB(IEADMCLx)'. Refer to the IEADMCLx
members for more information, or see
Table 1.
- For all types of logstreams, always include the following:
- The IXGLOGR (Logger) asid and the data spaces associated with
the IXGLOGR asid through the DSPNAME parm. These will be dumped using
the JOBNAME= parm.
- The trace data space SYSLOGR0, will be included in the dump if
DSNAME=('IXGLOGR'.*) is specified in the reply for the dump command.
Note: If
you are running OS/390® Release
2 or lower, Logger will not have a SYSLOGR* data space for tracing,
in which case the DSPNAME=('IXGLOGR'.*) option can be omitted.
DUMP COMM=(dump system logger with component trace)
* rr IEE094D SPECIFY OPERAND(S) FOR DUMP COMMAND
rr,JOBNAME=(IXGLOGR),CONT
ss,DSPNAME=('IXGLOGR'.SYSLOGR*),CONT
tt,SDATA=(COUPLE,ALLNUC,LPA,LSQA,PSA,RGN,SQA,SUM,
TRT,CSA,GRSQ,XESDATA),END
- When using CF list structure based log streams, include the following
- The XCF asid and trace data spaces. These will be dumped using
the JOBNAME= parm and DSPNAME parm.
- The XES STRUCTURE data. This is dumped using the STRLIST= parameter
and by specifying the structure name. structure_name is the affected
STRUCTURE name.
Note: Be sure to allocate adequate DUMPSPACE() as
defined in your CF definition in the CFRM policy. If you do not allocate
adequate space, all or part of the STRUCTURE will NOT be dumped.
- In the case of “loss of data” or “block not found”, dumping the
OFFLOAD data sets using IDCAMS is a good idea.
DUMP COMM=(dump system logger with xes and structure data)
* rr IEE094D SPECIFY OPERAND(S) FOR DUMP COMMAND
rr,JOBNAME=(IXGLOGR,XCFAS),CONT
ss,DSPNAME=('XCFAS'.*,'IXGLOGR'.SYSLOGR*),CONT
tt,SDATA=(COUPLE,ALLNUC,LPA,LSQA,PSA,RGN,SQA,SUM,
TRT,CSA,GRSQ,XESDATA),CONT
uu,STRLIST=(STRNAME=structure_name,LOCKENTRIES,ACC=NOLIM,
(LISTNUM=ALL,ADJUNCT=DIRECTIO,ENTRYDATA=UNSERIALIZE)),END
- When system logger dumps are needed on multiple systems in the
sysplex, include the REMOTE parameter.
DUMP COMM=(dumps for system logger around the sysplex)
* rr IEE094D SPECIFY OPERAND(S) FOR DUMP COMMAND
rr,JOBNAME=(IXGLOGR,XCFAS),CONT
ss,DSPNAME=('XCFAS'.*,'IXGLOGR'.SYSLOGR*),CONT
tt,SDATA=(COUPLE,ALLNUC,LPA,LSQA,PSA,RGN,SQA,SUM,
TRT,CSA,GRSQ,XESDATA),CONT
uu,STRLIST=(STRNAME=structure_name,LOCKENTRIES,ACC=NOLIM,
(LISTNUM=ALL,ADJUNCT=DIRECTIO,ENTRYDATA=UNSERIALIZE)),CONT
vv,REMOTE=(SYSLIST=*)'XCFAS','IXGLOGR'),DSPNAME,SDATA),END
- Other diagnostic considerations
The CTILOG00 buffer size is
16MB. It is strongly recommended that this value not be lowered. See
the CTnLOGxx member, described in Table 1.
Remember that much
of the data that system logger uses is persistent across an IPL. That
means that if this data is corrupted and adversely affects system
logger, an IPL will not correct the problem. In the case of a persistent
system logger failure, you can FORCE the IXGLOGR address space. Prior
to doing this you should bring down all of the applications connected.
Then issue the FORCE command (FORCE IXGLOGR,ARM)
and restart system logger using the supplied procedure in SYS1.PROCLIB
(IXGLOGRS). (S IXGLOGRS)
If FORCE IXGLOGR,ARM does
not resolve the situation, an IPL is not likely to either. This is
the time to take a dump if one was not already taken by system logger.
Note: - A CICS® dump will not dump
the IXGLOGR address space. Connect to a new (non-corrupted) LOGSTREAM.
This will result in a LOSS OF DATA for some applications such as CICS, forcing them to INITIAL START.
However, this may be the only way to get the application restarted.
Connecting to a new logstream (of a different name) will allow the
corrupted data to remain in the structure for diagnostic review later.
- In preparation for connecting to this new LOGSTREAM, you may want
to define an unused LOGSTREAM to each STRUCTURE during setup.
If running CICS,
always run with the following SLIP:
SLIP SET,COMP=1C5,REASON=804,
STRLIST=(STRNAME=strname1,LOCKE,ACC=NOLIM,
(LNUM=ALL,EDATA=SER,ADJ=CAP)),
JL=(XCFAS,IXGLOGR),DN=("XCFAS".*,"IXGLOGR".*),
SD=(COUPLE,ALLNUC,LPA,LSQA,PSA,RGN,SQA,TRT,CSA,GRSQ,
XESDATA,SUM),
SUMLIST=(13R?-7FFF,13R?+7FFF),END
Note: You might
add a JOBNAME=DFH* to limit SLIP to CICS.
A RSN804 is a “block not found”, which is always bad for CICS but not necessarily so for other applications.
Setting this SLIP will cause system logger to dump on all RSN804s
in CICS.
- Frequent stumbling blocks
- OFFLOAD data sets must be VSAM SHAREOPTIONS(3,3) unless you are
in a MONOPLEX.
- After OW33261, system logger recommends for performance reasons
using 24K CI size for OFFLOAD data sets. Staging data sets must remain
at 4K CI size. Code your ACS routines appropriately.
- Size of XES structures. “Bigger is not always better.” Follow
exploiting application recommendations.
- Allow for OFFLOAD directory extents. Reference “Format Utility
for Couple Data Sets” in z/OS MVS Setting Up a Sysplex.
- System logger uses HSM services to recall (ARCHRCAL) and to delete
(ARCHDEL) offload data sets. HSM contention or a wait for a WTOR such
as ARC0055A can hang all of the log streams that require the system
logger allocation task.