bcg365
Active Newcomer
Been searching for an answer and finding nothing in docs or from IBM support to confirm or deny, hoping someone may have inquired and either has or can point to a source for the answer.
DB2 backup through TSM working just fine. Issue is the databases are huge - 2TB+. This question is explicitly about object size due to post-processing actions on those objects, not the actual backup process itself (although they are naturally related).
The desire is to get smaller API objects, so that later activities such as backup to copy pools, exports, reclamations, etc. can work with smaller objects. Since each object holds a transaction for the duration of its processing, there are log pin issues on TSM <6 (which we still have some of - takes forever to migrate at our scale). In all TSM versions better parallelism can be achieved if the objects are smaller, even though ultimately the total data is processed anyway. One can reduce the concern with very clever storage pool layout, but it never quite goes away and as an Admin I'd rather use more straightforward means than always relying on clever tricks.
The API object size can be governed to some degree by using more sessions - assuming the target devices are equal, each session will roughly split the total data. However, my DBAs tell me that 8 sessions is the max - haven't found a manual or other IBM reference to confirm that either - so a 2TB database still generates 250GB objects. Also, sessions are simultaneous, so more sessions implies need for much bigger pipes or more target devices for LANFree, etc. (most of our backups use backplane network I/O in big shared LPAR pSeries frames, so there is bandwidth but it isn't infinite).
Target is to achieve database objects around 100GB. Easily done in Oracle as there is a parameter to RMAN to set the max size within a session after which RMAN opens a new object. TSM parameter txnbytelimit doesn't force a client (API or otherwise) to stop and start a new object, so that isn't a direction to go either (not that it would deal with sizes to 100GB).
I'm looking for a DB2 control parameter that would force the DB2 backup API through TSM to limit the object size within a session. This is probably more a DB2 question that a TSM question, but having had no luck from my DB2 resources (including IBM, who can't even say "you just can't") I'm hoping someone here may have tried to find and could share any insight.
Thanks in advance!
DB2 backup through TSM working just fine. Issue is the databases are huge - 2TB+. This question is explicitly about object size due to post-processing actions on those objects, not the actual backup process itself (although they are naturally related).
The desire is to get smaller API objects, so that later activities such as backup to copy pools, exports, reclamations, etc. can work with smaller objects. Since each object holds a transaction for the duration of its processing, there are log pin issues on TSM <6 (which we still have some of - takes forever to migrate at our scale). In all TSM versions better parallelism can be achieved if the objects are smaller, even though ultimately the total data is processed anyway. One can reduce the concern with very clever storage pool layout, but it never quite goes away and as an Admin I'd rather use more straightforward means than always relying on clever tricks.
The API object size can be governed to some degree by using more sessions - assuming the target devices are equal, each session will roughly split the total data. However, my DBAs tell me that 8 sessions is the max - haven't found a manual or other IBM reference to confirm that either - so a 2TB database still generates 250GB objects. Also, sessions are simultaneous, so more sessions implies need for much bigger pipes or more target devices for LANFree, etc. (most of our backups use backplane network I/O in big shared LPAR pSeries frames, so there is bandwidth but it isn't infinite).
Target is to achieve database objects around 100GB. Easily done in Oracle as there is a parameter to RMAN to set the max size within a session after which RMAN opens a new object. TSM parameter txnbytelimit doesn't force a client (API or otherwise) to stop and start a new object, so that isn't a direction to go either (not that it would deal with sizes to 100GB).
I'm looking for a DB2 control parameter that would force the DB2 backup API through TSM to limit the object size within a session. This is probably more a DB2 question that a TSM question, but having had no luck from my DB2 resources (including IBM, who can't even say "you just can't") I'm hoping someone here may have tried to find and could share any insight.
Thanks in advance!
