M204.0843

M204.0843  token This is an informational checkpoint message, which is written to both the audit trail and to the operator. The message is about timeout, abort, or completion. If it is about timeout, it will be one of these:

• Basic message format:

  Checkpoint timed out on yy.ddd hh:mm:ss.th due to user nn

• Sub-transaction checkpoint time out message:

  Sub-tran CP timed out on yy.ddd hh:mm:ss.th due to user nn

• An alternate form if additional information is available:

  Checkpoint timed out on yy.ddd hh:mm:ss.th due to user nn userid running procname(...)

  Sub-tran CP timed out on yy.ddd hh:mm:ss.th due to user nn userid running procname(...)

• The following form can only occur when the DKUPDTWT parameter is greater than zero:

  Checkpoint timed out on yy.ddd hh:mm:ss.th updating file filename

If it is about completion, the message is one of these:

Checkpoint completed on yy.ddd hh:mm:ss.th

Sub-transaction CHKP ID: nnnnn hh:mm:ss.th RCV=nnn

If it is about abort, the message is:

Sub-transaction checkpoint attempt aborted by user: nn

In these messages:

• yy.ddd is the Julian date.
• hh:mm:ss.th is time in hours, minutes, seconds, and tenths/hundredths of seconds.
• RCV=nnn is the number of blocks needed for sub-transaction recovery. The size of the recovery checkpoint dataset must be greater than the sum of the last RCV value plus the number of records in RESTART and RESTARTS.

System manager response: Checkpoint time-outs can be caused by one or both of the following problems:

1. The duration of typical update units is long compared to the CPTO or CPTQ parameters. Update units that span terminal I/O are a common example.
2. There are a very large number of modified pages in the buffer pool at the time a checkpoint is attempted, and it takes more than CPTO or CPTQ seconds to write them all to disk.

The first case should be addressed by reviewing update unit logic for the presence of COMMIT statements prior to terminal I/O statements. If that is not the problem, then it may be necessary to increase one of the parameters or to revise the application's update strategy.

The second case may be indicated by the occurrence of the FILE version of the timeout message. If this message is seen frequently, it may indicate that the Anticipatory Write Window is too small. The difference between the number of buffers allocated in the run and the setting of the LDKBMWND parameter controls the maximum number of unwritten modified buffers that can be in the buffer pool at a given time. If this message appears frequently, and time-outs cannot be traced to a specific, poorly behaving program, then consider increasing LDKBMWND.

Note: It is possible for the FILE version of the timeout message to appear, even though the problem is not the Anticipatory Write Window. If the typical duration of update units is very close to the setting of CPTO or CPTQ, there will be insufficient time to write even a small number of modified pages to disk.

Message attributes:

RETCODEO=0	Sets online return code
RETCODEB=0	Sets batch (single user) return code
CLASS=I	Information class; the message can be suppressed with the X'02' bit setting of the MSGCTL parameter
AUDITMS	Writes the message with line type MS to the audit trail
NOCOUNT	Does not increment the error count (ERCNT) parameter
NOTERM	Does not display the message on the user's terminal
OPR	Writes the message to the (operator) console

Back to list of messages