[improve][storage] Periodically rollover Cursor ledgers

[eolivelli]

Motivation

Each ManagedCursor stores its own status on a BookKeeper ledger and this ledger is rolled over when one of these conditions are met: some time passed or too many entries have been written.

The problem is that the condition is checked only while writing to the ledger: if no consumer sends acknowledgments then the ledger is never rolled over.
This may become a problem because there will be a long lived ledger that doesn't respect the configuration set by the administrator for "ledger rollover".
Creating a new ledger allows the Bookies to reclaim space, especially thanks to the GC (Garbage collection) mechanism, and then with the Compaction.
Having a long lived ledger prevents the Bookie to release disk space promplty with the GC, and so only Compaction can release space, but compaction is an expensive operation.
For this reason in some usecases you want to configure a frequent rollover of the ledgers for the topics, and thus you want that such rollover happens at the same frequency also for the cursor ledgers.

Modifications

This patch simply allows the broker to periodically check every ManagedCursor and apply the rollover configuration even when there is no consumer activity.

Verifying this change

This change added tests

Does this pull request potentially affect one of the following parts:

If the box was checked, please highlight the changes

• Dependencies (add or upgrade a dependency)
• The public API
• The schema
• The default values of configurations
• The threading model
• The binary protocol
• The REST endpoints
• The admin CLI options
• The metrics
• Anything that affects deployment

Documentation

• doc
• doc-required
• doc-not-needed
• doc-complete

Matching PR in forked repository

PR in forked repository: eolivelli#26

[thetumbled]

LGTM.
I agree that we should find out the long live ledgers and shorten their lifecycle, so that we do not need to speed lots of effort on compaction.
According to my finding, the space reclaimed by compaction take up a large part of the total space reclaimed, which will spend lots of I/O resource. We should increase the ratio of space reclaimed by deleting the entire entry log.

[michaeljmarshall]

LGTM

[lhotari]

LGTM