Status Variables

SHOW GLOBAL STATUS LIKE 'tidesdb%';

These are the machine-readable counters for a monitoring agent such as a Prometheus exporter or PMM. Monitoring explains which of them matter and what healthy looks like. They are refreshed on demand behind a short coalescing window, so reading many of them in one statement costs a single stats pass.

Identity

Variable	Description
`Tidesdb_version`	TideSQL plugin version string, for example `5.2.1`
`Tidesdb_version_hex`	Plugin version as an integer, for example `328193` for `0x50201`
`Tidesdb_library_version`	Version string of the TidesDB library the plugin was built against, for example `10.2.1`

Sequence and transactions

Variable	Description
`Tidesdb_column_families`	Number of active column families
`Tidesdb_global_sequence`	Global MVCC sequence number
`Tidesdb_min_snapshot_sequence`	Oldest pinned snapshot, the floor compaction cannot reclaim past
`Tidesdb_active_transactions`	Transactions currently joined to the MVCC registry. `READ UNCOMMITTED` transactions do not join it and are not counted
`Tidesdb_txn_memory_bytes`	Memory held by in-flight transactions in bytes

Transaction errors

Counters bumped each time a library error is mapped to a server error. A failure at COMMIT counts once in Tidesdb_commit_conflicts whatever its cause, and a failure inside a statement counts in the Tidesdb_stmt_* counter for its cause, so real contention can be told apart from engine churn. These are read live rather than through the refresh window.

Variable	Description
`Tidesdb_commit_conflicts`	COMMIT statements that failed with a deadlock or lock wait timeout error. A failure in the prepare phase, which the server runs when the binary log is on, counts here too. The server counts a failed COMMIT in `Com_commit` too, so subtract this to get successful commits
`Tidesdb_stmt_conflicts`	Statements that failed on a write conflict, or because the library ran out of retries under heavy memtable rotation, returned as a deadlock error
`Tidesdb_stmt_locked`	Statements that failed on transient contention the library reports as locked, returned as a lock wait timeout after a short bounded retry
`Tidesdb_stmt_memory_limit`	Statements that failed on memtable backpressure, returned as a lock wait timeout
`Tidesdb_stmt_txn_expired`	Statements that failed because the transaction outlived `tidesdb_txn_timeout_seconds`, returned as a lock wait timeout
`Tidesdb_stmt_bf_aborted`	Statements that failed because a Galera applier brute-force aborted the transaction, returned as a deadlock error

Memory and storage

Variable	Description
`Tidesdb_memtable_bytes`	Memory the active memtable occupies in bytes, skip-list nodes and version structs included alongside key and value bytes. This is the figure compared against `tidesdb_memtable_write_buffer_size`
`Tidesdb_total_sstables`	Total SSTable count across all column families
`Tidesdb_open_sstables`	Open SSTable file handles
`Tidesdb_data_size_bytes`	On-disk key-log bytes summed across every column family and level. The value log is reported separately in `Tidesdb_vlog_file_size`, and the WAL is not included
`Tidesdb_immutable_memtables`	Sealed memtables waiting to be flushed
`Tidesdb_flush_pending`	Flushes pending (the immutable memtable queue depth)
`Tidesdb_compaction_queue`	Compaction jobs queued for the worker pool
`Tidesdb_memtable_is_flushing`	1 while an immutable is queued or flushing
`Tidesdb_wal_generation`	Current write-ahead-log generation counter

Value log

Variable	Description
`Tidesdb_vlog_file_size`	On-disk size of the value log in bytes
`Tidesdb_vlog_value_count`	Values the value log currently indexes
`Tidesdb_vlog_used_bytes`	Uncompressed length those values represent
`Tidesdb_vlog_bytes_written`	Lifetime bytes appended to the value log, output the flush and compaction counters do not see once values separate

Encoding

Aggregate codec-chain totals summed across every chain, so logical divided by stored is the realized compression ratio for each log. SHOW ENGINE TIDESDB STATUS prints the same totals per log with the chain count and the ratio.

Variable	Description
`Tidesdb_klog_logical_bytes`	Key-log bytes before encoding, summed across chains
`Tidesdb_klog_stored_bytes`	Key-log bytes after encoding as stored on disk
`Tidesdb_vlog_encoded_logical_bytes`	Value-log bytes before encoding, summed across chains
`Tidesdb_vlog_encoded_stored_bytes`	Value-log bytes after encoding as stored on disk

Device IO

Write accounting from the library’s file-descriptor manager, which meters SSTable, WAL, and value-log writes. These counters surface the SSTable and WAL classes, the value log’s write volume is in Tidesdb_vlog_bytes_written above, and all three classes print with their write timing in SHOW ENGINE TIDESDB STATUS. These counters are writes only, there is no read-side figure.

Variable	Description
`Tidesdb_io_sstable_write_ops`	SSTable device writes issued since open
`Tidesdb_io_sstable_write_bytes`	Bytes written to the SSTable device since open
`Tidesdb_io_wal_write_ops`	WAL device writes issued since open
`Tidesdb_io_wal_write_bytes`	Bytes written to the WAL device since open

Write amplification

Variable	Description
`Tidesdb_user_bytes_written`	Logical committed bytes, the write-amplification denominator
`Tidesdb_flush_bytes_written`	Bytes written to SSTables by flush jobs
`Tidesdb_compaction_bytes_written`	Bytes written by compaction jobs
`Tidesdb_compaction_bytes_read`	Bytes compaction read as input
`Tidesdb_flush_count`	SSTables written by flushes, one per column family per flush
`Tidesdb_compaction_count`	Compaction jobs completed across all column families. It counts jobs, not the SSTables they write

Write stalls

Variable	Description
`Tidesdb_writes_throttled`	Commits the L0 admission policy made dwell before admitting
`Tidesdb_writes_blocked`	Commits it made wait for the flush queue or the level-1 backlog to drain
`Tidesdb_write_stall_us`	Total microseconds commits spent held in admission
`Tidesdb_write_stall_ceiling_hits`	Commits admitted only because the wait ceiling expired. Any sustained increase means flush is not keeping up with ingest

The aggregate counters above sum admission stall time. These per-reason counts split how often a commit stalled by cause, so backpressure can be attributed. The per-reason stall time prints in SHOW ENGINE TIDESDB STATUS.

Variable	Description
`Tidesdb_stall_wal_append`	Commits that stalled appending to the write-ahead log
`Tidesdb_stall_rotate_lock`	Commits that stalled taking the memtable rotation lock
`Tidesdb_stall_rotate_work`	Commits that stalled while a memtable rotation was in progress
`Tidesdb_stall_admission`	Commits that stalled on the unflushed-backlog admission gate
`Tidesdb_stall_manifest_commit`	Commits that stalled waiting for a manifest commit

Block cache

Variable	Description
`Tidesdb_cache_entries`	Cached entry count
`Tidesdb_cache_bytes`	Bytes used by the block cache
`Tidesdb_cache_hits`	Cache hits since open
`Tidesdb_cache_misses`	Cache misses since open
`Tidesdb_cache_hit_rate`	Hit rate as a percentage
`Tidesdb_cache_partitions`	Number of cache shards

Tombstones

Variable	Description
`Tidesdb_total_tombstones`	Total tombstones summed across every column family
`Tidesdb_tombstone_ratio`	Database-wide tombstone count divided by entry count, 0.0 to 1.0
`Tidesdb_max_sst_tombstone_density`	Worst single-SSTable tombstone density observed
`Tidesdb_max_sst_tombstone_density_level`	1-based LSM level where the worst SSTable sits, `0` if none was observed