Skip to main content

WAL Diagnostics

Kommander exposes WAL visibility helpers and a practical durability switch for test environments.

WAL Batch Metrics​

The WAL scheduler reports batching through:

  • raft.wal.batches_total
  • raft.wal.operations_total
  • raft.wal.batch_size

raft.wal.batches_total counts scheduler group writes to the storage adapter. A single group write can include work from multiple partitions.

raft.wal.batch_size records the number of operations drained for each partition inside a group write. It does not report how many partitions were included in the group.

Under load, a healthy batching pattern usually shows raft.wal.operations_total growing faster than raft.wal.batches_total, which means multiple WAL operations are being amortized into fewer storage calls.

Counting Persisted And Removable Logs​

IWAL exposes two helpful counting methods:

  • CountPersistedLogs(partitionId)
  • CountRemovableLogs(partitionId)

CountPersistedLogs returns the total number of persisted log rows for the partition.

CountRemovableLogs returns the number of persisted rows strictly below the last committed checkpoint.

These are useful for:

  • compaction diagnostics
  • WAL-focused tests
  • operational checks that want to estimate how much history is still removable.

Optional Non-Synchronous Writes​

Both durable adapters support syncWrites: false:

IWAL rocks = new RocksDbWAL("./data", "node-1", logger, syncWrites: false);
IWAL sqlite = new SqliteWAL("./data", "node-1", logger, syncWrites: false);

This can improve throughput in CI, benchmarks, and some local test runs.

Durability Tradeoff​

With syncWrites: false, acknowledged writes may still be lost on process or machine crash. Use it only when crash durability is not part of what you are validating.