pulsar

<@UQ8NL176Y> <@UGFSX2BNK> <@U039LGT7V3J>
Could you guys help answer this question if possible by any chance? Thanks!

don’t delete anything in ledger data dir. you should use bookie shell command to compact/gc the entry log.

<https://bookkeeper.apache.org/docs/4.9.2/getting-started/concepts/#data-compaction>

<https://bookkeeper.apache.org/docs/4.9.2/getting-started/concepts/#entry-logs>

Thanks for your help. Apparently the ledger disk was not completely full (95% then bookie became readonly) but then /var disk got full but seemed to be cleaned up together with ledgers. Gc was attempted via http api but hardly made any space, compaction endpoint is not available in our bookkeeper version but we will update soon.

Backstory of issue: we had some ephemeral environments for which pulsar namespaces were created without ttl/retention policies/backlog quotas

you can add 1 new bookie and use autorecovery and decommission the read-only to force the ledger moving to new node.