Using symmetrical clusters

A symmetrical cluster ensures that IBM® Notes® databases remain identical across all servers in a cluster. A repair service repairs missing or damaged databases by replacing them with good copies from donor cluster members.

For symmetrical cluster requirements, see the topic Preparing for a symmetrical cluster.

Detection and repair of missing databases
A server task, AutoRepair, runs on each server in a symmetrical cluster. AutoRepair scans specified folders on the server for missing databases. AutoRepair refers to the Cluster Database Directory (cldbdir.nsf) to determine which databases the monitored folders should contain. When AutoRepair detects and repairs missing databases, it also repairs any DAOS .nlo files that are referenced by the databases.
You use a Cluster Configuration document to configure the following aspects of AutoRepair:
  • Whether AutoRepair only reports missing databases to the log file or whether it triggers the repair service to replace missing databases.
  • The folders to scan. Folders must be below the server data directory.
  • The frequency with which AutoRepair scans for missing databases.
Detection and repair of damaged databases
To detect and repair damaged databases, you select the Fix damaged files option in a Cluster Configuration document. When this option is selected and a Domino® server detects a damaged database in a monitored folder, Fixup runs to attempt to repair the database. If Fixup is unsuccessful, the server quarantines the damaged database by changing its file extension to .pd_bad_<timedate>. The repair service immediately recovers a good version from a donor cluster member. The server task RepairCleanup deletes the quarantined versions of databases after a number of days that you specify in the Cluster Configuration document.
Repair service
To replace a missing or damaged database, the repair service:
  • Copies a snapshot of the database from a donor server that has a good version.
  • Updates folder references.
  • Adjusts the replication history for cluster mate copies to reflect the point of divergence to minimize replication impact.
  • Rebuilds the full text index if full-text indexing is enabled on the database that was copied.
  • Scans the database to verify that any referenced DAOS objects are present on this server. If any are missing, it replaces them.