Best Practices

Bacula Enterprise Only

This solution is only available for Bacula Enterprise. For subscription inquiries, please reach out to sales@baculasystems.com.

The following article presents best practices for job distribution, concurrency, and performance.

Use one Bacula job per IRIS instance when possible. This reduces contention and makes restore selection easier.
Prefer Online Backup for most environments. Use External/Concurrent External only when storage snapshots are required.
Keep iris_owner_user and irissession paths correct and test them manually before scheduling backups.
When using selective restores, generate directory list files directly from IRIS tools to avoid typos.
Maintain a separate IRIS instance for restore validation in production environments.
Avoid aggressive concurrent_threads settings unless you have verified sufficient CPU and disk bandwidth.
Set cleanup=yes only if you have confirmed that Bacula successfully stored the backup artifacts.

Concurrency Best Practices

Avoid overlapping backup jobs on the same instance; IRIS allows only one online backup at a time.
Stagger backup schedules for multiple instances on the same host to prevent I/O contention.
For high-load environments, start with conservative values (e.g., concurrent_threads=50) and increase gradually.
Ensure the File Daemon and Storage Daemon have sufficient throughput before raising concurrency parameters.

Test restores regularly on a non-production IRIS instance using the same Fileset and job definitions.
Validate both full restore and selective restore paths (directories_path and journal replay options).
Keep a documented runbook including required restore parameters and expected artifact locations.
For critical systems, retain a known-good full backup and a recent incremental chain to reduce RTO.

Use this checklist before increasing workload concurrency or backing up multiple instances on the same host:

CPU: verify adequate headroom during peak load; avoid raising concurrent_threads if CPU is already saturated.
Disk I/O: ensure the IRIS data volume and Bacula storage path can sustain backup throughput without latency spikes.
Network: validate sustained bandwidth between File Daemon and Storage Daemon for the expected backup window.
JVM Memory: increase IRIS_JVM_MAX if large backups or many parallel jobs risk OutOfMemory errors.
Scheduling: avoid overlapping backups for the same instance; stagger jobs for multiple instances.