pulsar_cluster log will fill up the disk

[1275718633]

The log space in the "ledgers/current" directory under the journalDirectory path configured in bookkeepr is too large and is about to fill up the disk. How should I handle this?

[lhotari]

This is a good question. Some quick hints to prevent this from happening:

• read https://pulsar.apache.org/docs/4.0.x/cookbooks-retention-expiry/ to understand the basics
• Pulsar will retain message backlogs on subscriptions, so you will need to monitor backlog sizes or set policies to prevent disk space over use
  • storage metrics at namespace or topic level: https://pulsar.apache.org/docs/4.0.x/reference-metrics/#topic-metrics

Some useful details to be aware of:

• Only fully consumed (no subscription and no other retention) ledgers for a topic can be removed
• BookKeeper won't release data from deleted ledgers until there's a BookKeeper garbage collection cycle
  • Garbage collection settings

For test environments, you can use use these settings for more aggressive disk cleanup

bookkeeper:
  configData:
    # more aggressive disk cleanup
    journalMaxSizeMB: "256"
    majorCompactionInterval: "600"
    minorCompactionInterval: "300"
    compactionRateByEntries: "5000"
    gcWaitTime: "60000"

broker:
  configData:
    # more aggressive disk cleanup
    managedLedgerMinLedgerRolloverTimeMinutes: "1"
    managedLedgerMaxLedgerRolloverTimeMinutes: "5"