• June 27, 2022

The fun with CPF4904 Maximum size for member reached

Every programmer and most operators have seen this message at some point in their job…CPF4903 The maximum size for member NEWMBR has been reached for file MYFILE in library MYLIB.


This is easily handled on the primary node by answering the inquiry message to add another increment, but if this happens on the backup node where there may not be anyone around to answer, this can cause some issues. Usually you will see latency in the replication group as the apply job HADTUP waits for an answer. You can handle this by adding a system reply list entry on the backup node to issue an automatic Ignore or by changing the file itself to have *NOMAX for the member size.

In iCluster 7.1 IF1 we added an enhancement request to have iCluster automatically change the file on the backup node to *NOMAX for maximum number of increments. Well that fixed the issue for some customers BUT some legacy S38 applications do not allow *NOMAX for their files. Ooops.

In TR2 we have now provided a method of optionally telling iCluster to keep the member size to exactly what it was on the primary. The method is very similar to our post last week on how to keep triggers enabled on the backup node. Manually change the metadata file ICLUSTER/DMGROUPS on both nodes by updating the RESERVED2 field to ‘*KEEPSIZE’ (keep member size) in the record for the group. You can use the command below:


After the DMGROUPS metadata files are updated ON BOTH NODES, end the group and issue a DMSETPOS then restart the group. Now you will have to go back to whatever method you were previously using to handle the old CPF4903.

This fix will be included in both iCluster 7.1 Interim Fix 12 and iCluster 7.1 TR1 Interim Fix 9 (both interim fixes are yet to be released), but is already available in iCluster 7.1 TR2!

iCluster Tech Tuesday 0 Posts

iCluster TechTuesday is a set of posts covering technical tips and techniques to help get the most out of your Rocket iCluster installation.


Leave a Comment

Your email address will not be published.