April 16 2015
What is required?
You must do a logical backup of the data set so the first page in the data set is backed up first.
If you do a volume backup, then if your data set is spread across multiple volumes,
or multiple extents on one volume it is possible that the first page in the data set
is not backed up first. If you need to restore from one of these bad backup, it will restore,
and appear to work, but you may get data inconsistency at a later date – perhaps weeks later.
If you are doing a backup while the queue manager is active ( a fuzzy backup) the duration of the backup does not matter.
For example using FastReplication within the DS88XX subsystem is fast. Backing up to a tape will be slower, as all of the data will need to be copied.
There are different ways of backing up data. They take different durations and CPU time.
When you use TOL(ENQF) to backup while the queue manager is active Concurrent Copy, Fast Replication and a straight ADRDSSDU produce usable fuzzy backups that can be recovered from.
You can use ADRDSSU from IBM to backup datasets such as MQ page sets.
We used to recommend using Concurrent Copy, but with modern DASD there are better ways ( which use less CPU) than concurrent copy. Concurrent copy allowed the data set to be backed up while minimizing the time the dataset was unavailable. This is done at the z/OS level.
These days, with PC sized disks instead of big spinning 3390 disks, there are better ways of backing up the dataset. For example the DS88xx subsystem can quickly copy a data set within the subsystem, by copying the pointers to the data, instead of copying the data itself. A data page is then pointed to by two datasets. As the page set is updated, the pointer then points to the new updated page.
The ADRDSSU parameter FastReplication controls how the DASD subsystem copies the dataset. The value are
- None – do not use Fast replication
- Required – It must use it
- Preferred – (the default) use it if available.
I did some measurements to backup a 3317 MB page set
|how||TCB CPU||SRB CPU|
If you are copying out of the DASD subsystem – such as to tape, then FastReplication(NO) will be used, and the all the data will be copied.
You should use ADRDSSU instead of VSAM Repro as it is much faster.
Here is the JCL I use
//STEP2 EXEC PGM=ADRDSSU,REGION=6M
//SYSPRINT DD SYSOUT=H
// SET BACKUP=SCENDATA.MQPA.PAGESET.P4.BACKUP
//SYSIN DD *
April 14 2015
MQ V701, V710 and V8 support SCM storage for CF structures.
System Class Memory (SCM) uses Flash Express cards which are Solid State Devices
to provide additional storage to z/OS. The CF can be configured to use this storage.
This is a z/OS level facility.
MQ’s shared queue can use this with no change, though there are some APARs in this area. You should check with IBM what APARs you need.
Beginning with CFLEVEL 19, you can configure storage-class memory to a coupling facility LPAR. Storage-class memory provides an overflow capability to minimize the probability of structure-full conditions. As with coupling facility real storage, you cannot dynamically reconfigure storage-class memory, nor can one partition
reclaim storage-class memory that is configured to another partition.
Use of storage-class memory increases the amount of coupling facility real storage that is required by the affected structures.
MQ can take advantage of this facility and avoid structure full condition in the peak of workload.
The configuration is within the CF and no additional MQ configuration is required.
If you intend to exploit SCM, ie. you have structures at CFLEVEL 19 with SCMMAXSIZE and SCMALGORITHM specified in the CFRM
policy, you should apply:
PI13906 PTF UI19595 HIPER
PI21909 PTF UI20147
PI13080 PTF UI18181
CF structures allocated in CFLEVEL 19 CF may need to be significantly larger in order to be allocated with a similar number of usable structure objects.
It is highly recommended to use the CFSIZER tool:
Some additional information available
Enhancements for z/OS Parallel Sysplex
IBM MQ V8 Features and Enhancements chapter 8
These blog posts are from when I worked at IBM and are copyright © IBM 2015.