WAL and Durability

The write-ahead log (WAL) is the recovery bridge between an append call and a compressed segment block.

When an append is accepted, TSEdge records it in the WAL before relying on the in-memory buffer. If the process crashes before the buffer is flushed to a segment, the next tsedge_open can replay complete WAL records.

Recovery behavior

• Complete WAL records are replayed on database open.
• An incomplete final WAL record is treated as a torn write and ignored.
• A complete record with a bad checksum is treated as corruption.
• WAL recovery is local to the embedded database directory.

TSEdge WAL is not a full transactional database engine. It restores accepted records but does not provide multi-series transactional atomicity.

Durability modes

FAST

FAST favors throughput. WAL data may be buffered before reaching wal.log, so a process or system crash can lose the most recent buffered appends.

BALANCED

BALANCED flushes WAL data more often than FAST while avoiding the cost of syncing every append.

STRICT

STRICT flushes WAL data on every append or batch append. It provides the strongest crash recovery behavior available in TSEdge.

Flush and close

tsedge_flush, tsedge_flush_all and tsedge_close flush pending data through the normal WAL and segment path before completing. For most applications, closing the database cleanly is enough to persist the current buffers into segment files.