This test passes consistently when the test suite is run with a
single thread. It fails consistently on MacOS when run as part
of the unit-test suite.
No idea why it passes in CI.
By passing a config and not a Arc'ed LeaderScheduler, callers
need to use `Bank::leader_scheduler` to access the scheduler.
By using the new constructor, there should be no incentive to
reach into the bank for that object.
* Modify db_ledger to support per_slot metadata, add signal for updates, and add chaining to slots in db_ledger
* Modify replay stage to ask db_ledger for updates based on slots
* Add repair send/receive metrics
* Add repair service, remove old repair code
* Fix tmp_copy_ledger and setup for tests to account for multiple slots and tick limits within slots
* Modify replay stage to ask db_ledger for updates instead of reading from upstream channel
* Add signal for db_ledger to update listeners about updates
* fix flaky test
The bpfloader crate was triggering cargo to perform excessive rebuilds
of in-workspace dependencies. Unclear why exactly, but seems related to
the special dual crate-type employed by bpfloader.
* Groundwork for entry tree, align constants with definitions in the book
* Fix edge case in test, node can keep generating ticks between handle_role_transition and exit() call
* Connect TPU's broadcast service with TVU's blob fetch stage
- This is needed since ledger is being written only in TVU now
* fix clippy warnings
* fix failing test
* fix broken tests
* fixed failing tests
Split up StatusDeque into different modules
* LastIdQueue tracks last_ids
* StatusCache keeps track of signature statuses
* StatusCache stores success as a bit in a bloom filter
* Overhead for 1m Ok transactions is 4mb in memory
* Less concurrency between the objects, last_id and status_cache are read and written to at different points in the pipeline
* Each object has its own strategy for merging into the root checkpoint
