Repository Manager

In a single queue manager estate, the local object repository is enough: DEFINE QLOCAL on QM1 stays on QM1. Add fifty queue managers in a cluster and manual replication becomes impossible. The repository manager component keeps cluster object knowledge consistent—publishing new cluster queue definitions, updating CLUSRCVR records, and helping partial members learn routes. When it falls behind, applications still put messages but workload may target stale instances or channels fail to auto-define. This tutorial explains full versus partial repositories, publication flow, REFRESH CLUSTER, common inconsistency symptoms, and how the repository manager differs from the everyday repository covered in earlier pages.

Cluster Repository Roles Recap

Designate one or two full repository queue managers per cluster for resilience. All other cluster members are partial repositories. The repository manager on a full repo receives publishes from members when they DEFINE or ALTER cluster objects and propagates updates. Partial repository managers request and cache definitions they need for routing puts to cluster queues and for auto-channel creation.

Explainer: Publication When You DEFINE CLUSQM

When you define a cluster queue or alter CLUSRCVR on a member, the change is published to full repositories. Repository managers merge updates and fan out to interested partial members. DISPLAY CLUSQ displays cluster queue attributes aggregated from repository knowledge. Delayed publication shows up as partner queue managers unaware of a new instance until refresh completes—puts may still work via previous default instance until CLWL chooses anew.

REFRESH CLUSTER and Recovery

After major maintenance—restoring a full repository from backup, fixing corrupted cluster definitions, or joining a new data center—operators run REFRESH CLUSTER with appropriate options (repository, security, or both per command syntax in your version). This forces repository manager resynchronization. Schedule during change windows; bursts of control traffic may occur. Pair with verification DISPLAY CLUSQ and TEST CHANNEL checks.

Troubleshooting Inconsistencies

• Compare DISPLAY CLUSQMGR on multiple members for missing attributes.
• Verify cluster sender/receiver channels to full repositories are RUNNING.
• Check for manual DELETE of cluster objects on one node only.
• Review AMQERR for repository manager messages after network splits.
• Ensure two full repositories do not diverge—maintain symmetric admin procedures.

Relation to Local Repository

The repository of objects tutorial described DEFINE and DISPLAY on one queue manager. The repository manager does not replace that—each queue manager still persists its local definitions. Cluster information may appear as cluster cache entries and mirrored definitions on partial repos. dmpmqcfg on one node exports local view; cluster-wide truth may require exports from full repository managers.

Tutorial: Inspect Cluster Repository Health

1
2
3
4
5
6
7
DISPLAY QMGR CLUSTER
DISPLAY CLUSQMGR(*) WHERE(CLUSTER EQ 'MYCLUSTER')
DISPLAY CLUSQ('ORDERS.IN') CLUSTER('MYCLUSTER')
* On full repository QM:
REFRESH CLUSTER MYCLUSTER REPOS
* Verify channel to repo QM:
DISPLAY CHSTATUS('TO.REPO') ALL

Explain Like I'm Five: Repository Manager

Many post offices in a chain share one master list of which towns have which mailboxes (cluster repository). The repository manager is the worker who updates every branch's copy when a new mailbox opens. If the master list is wrong, letters go to the wrong town.

Practice Exercises

Exercise 1

Full repo QM_A failed; QM_B is backup full repo. What must channels do before REFRESH?

Exercise 2

New cluster queue not visible on spoke. List four checks involving repository manager.

Exercise 3

Explain difference between repository manager and command server.

Frequently Asked Questions

Test Your Knowledge