A cluster is a named group of queue managers that publish their queues to each other. An application puts to ORDERS.IN. The cluster repository knows which queue managers host that name, and a cluster channel moves the message. You do not define a QREMOTE on every sender for every destination. Naming standards still matter: the cluster name, the queue name, and the channel names are what the repository stores.
| Piece | Role |
|---|---|
| Cluster name | The string on CLUSRCVR, CLUSSDR, and CLUSTER() |
| Full repository | Holds the directory. Run two. |
| Partial repository | Caches what it looks up |
| Cluster queue | QLOCAL with CLUSTER(name) |
| Cluster channels | CLUSSDR to a full repository, CLUSRCVR for inbound |
123456DEFINE QLOCAL('ORDERS.IN') REPLACE + CLUSTER('PROD.ORDERS') + DEFPSIST(YES) + MAXDEPTH(200000) + DESCR('Orders intake - shared across PROD.ORDERS') DISPLAY QCLUSTER('ORDERS.IN')
CLUSTER('PROD.ORDERS') is the advertisement. The queue manager must already be a member of that cluster, which means its cluster-receiver channel and its sender toward a full repository use the same name. A queue defined only as QLOCAL, with no CLUSTER attribute, stays private to that queue manager. Other members will not see it, and a put from elsewhere returns 2085 for that name.
Two full repositories are the usual production layout, on different hosts. Every other queue manager is partial: it asks a full repository when it needs a route and keeps a cache. Losing one full repository leaves the other. Losing both leaves members with only what they have already cached.
Each post office has its own mailboxes. The cluster is one shared phone book that says which office holds mailbox ORDERS.IN. You address the letter to the mailbox name. The phone book and the delivery vans (channels) pick the office. You do not draw a private map to every building.
Define ORDERS.IN in cluster PROD.ORDERS and say which DISPLAY shows the cluster view.
A put from QM_A returns 2085 for a queue that exists on QM_B. Name the attribute that was probably left off the DEFINE.
Say why the cluster keeps two full repositories.
1. A clustered queue is advertised with:
2. Production clusters usually run:
3. A partial repository:
4. Two queue managers hosting the same queue name can: