5.3. Metrics reference

This reference lists the kernel metrics that QuasarDB reports. The available metrics can depend on the QuasarDB version and configuration.

A metric has one of these types:

• An accumulator is an integer that increases during the life of the process.

• A gauge is an integer that can increase or decrease.

• A label is a text value.

5.3.1. Cache

These metrics describe the QuasarDB cache and its two-level LRU2 metadata cache. A new LRU2 entry is cold. QuasarDB promotes an entry to the hot cache after another access.

Metric name	Type	Description	Notes
evicted.count	accumulator	The total number of entries evicted from the QuasarDB cache.	Deprecated.
evicted.total_bytes	accumulator	The total number of bytes evicted from the QuasarDB cache.	Deprecated.
lru2.cold.count	gauge	The current number of entries in the cold LRU2 metadata cache.
lru2.cold.evicted.count	accumulator	The total number of entries evicted from the cold LRU2 metadata cache.
lru2.cold.evicted.total_bytes	accumulator	The total number of bytes evicted from the cold LRU2 metadata cache.	The byte value is not exact.
lru2.cold.pagein.count	accumulator	The total number of entries read into the cold LRU2 metadata cache.
lru2.cold.pagein.total_bytes	accumulator	The total number of bytes read into the cold LRU2 metadata cache.	The byte value is not exact.
lru2.hot.count	gauge	The current number of entries in the hot LRU2 metadata cache.
lru2.hot.evicted.count	accumulator	The total number of entries evicted from the hot LRU2 metadata cache.
lru2.hot.evicted.total_bytes	accumulator	The total number of bytes evicted from the hot LRU2 metadata cache.	The byte value is not exact.
lru2.hot.hit.count	accumulator	The total number of hits in the hot LRU2 metadata cache.
lru2.hot.hit.total_bytes	accumulator	The total number of bytes returned by hits in the hot LRU2 metadata cache.	The byte value is not exact.
lru2.hot.promoted.count	accumulator	The total number of entries promoted from the cold cache to the hot LRU2 metadata cache.
lru2.hot.promoted.total_bytes	accumulator	The total number of bytes promoted from the cold cache to the hot LRU2 metadata cache.	The byte value is not exact.
pageins.count	accumulator	The total number of entries that QuasarDB read from disk into its cache.	Deprecated.
pageins.total_bytes	accumulator	The total number of bytes that QuasarDB read from disk into its cache.	Deprecated.

5.3.2. Memory

These metrics describe memory use by the host, the QuasarDB process, RocksDB, and the TBB allocator.

Host, resident-cache, and persistence memory gauges are part of the normal statistics snapshot. The periodic memory.tbb.* allocator gauges require additional instrumentation, which is disabled by default. Set local.limiter.memory_statistics_refresh to an interval of at least 1,000 milliseconds to enable those gauges. A value of 0 leaves the allocator instrumentation disabled. See Configuration for all forms of the option.

Metric name	Type	Description	Notes
memory.persistence.cache_bytes	gauge	The number of bytes in all RocksDB block caches. QuasarDB does not use the RocksDB row cache.
memory.persistence.memtable_bytes	gauge	The number of bytes in the RocksDB memtables.
memory.persistence.memtable_unflushed_bytes	gauge	The number of unflushed bytes in the RocksDB memtables.
memory.persistence.table_reader_bytes	gauge	The number of bytes used by all RocksDB table readers.
memory.persistence.total_bytes	gauge	The total number of bytes used by RocksDB memtables, table readers, and block caches.
memory.physmem.total_bytes	gauge	The total number of physical-memory bytes detected on the host.
memory.physmem.used_bytes	gauge	The number of physical-memory bytes used on the host.
memory.resident_bytes	gauge	The number of bytes used by entries that are in memory.
memory.resident_count	gauge	The number of entries in the internal in-memory hash table. This value is related to, but not equal to, the table count.
memory.tbb.global_loc_total_bytes	gauge	An internal TBB allocator value used to diagnose complex memory problems.
memory.tbb.huge_threshold_bytes	gauge	The allocation threshold, in bytes, at which TBB uses huge pages when the system supports them.
memory.tbb.large_object_bytes	gauge	The number of bytes used by large objects that do not fit in the optimized allocation structures.
memory.tbb.large_object_count	gauge	The number of large objects that do not fit in the optimized allocation structures.
memory.tbb.large_unaligned_bytes	gauge	An internal TBB allocator value used to diagnose complex memory problems.
memory.tbb.max_requested_bytes	gauge	The largest number of bytes in one allocation request.
memory.tbb.softlimit_bytes	gauge	The threshold, in bytes, at which TBB returns memory to the operating system. Below this threshold, TBB keeps the memory.
memory.tbb.thread_loc_total_bytes	gauge	An internal TBB allocator value used to diagnose complex memory problems.
memory.tbb.thread_small_allocated_bytes	gauge	An internal TBB allocator value used to diagnose complex memory problems.
memory.tbb.thread_small_bin_count	gauge	An internal TBB allocator value used to diagnose complex memory problems.
memory.tbb.thread_small_block_count	gauge	An internal TBB allocator value used to diagnose complex memory problems.
memory.tbb.thread_small_count	gauge	An internal TBB allocator value used to diagnose complex memory problems.
memory.tbb.thread_small_used_bytes	gauge	An internal TBB allocator value used to diagnose complex memory problems.
memory.tbb.total_bytes	gauge	The number of bytes currently managed by TBB. TBB does not manage all QuasarDB allocations.
memory.tbb.total_count	gauge	The total number of allocations made through TBB.
memory.vm.total_bytes	gauge	The total number of virtual-memory bytes available to the process. This value is usually very large on a 64-bit operating system.
memory.vm.used_bytes	gauge	The number of virtual-memory bytes used by the process. Reserved memory can make this value larger than physical-memory use.

5.3.3. Clustering

These metrics describe the cluster topology and cluster synchronization.

Metric name	Type	Description	Notes
chord.invalid_requests_count	accumulator	The total number of requests that clients sent to an incorrect node.
chord.node_id	label	The node identifier.
chord.predecessor_changes_count	accumulator	The total number of predecessor changes. Frequent changes can identify a cluster problem.
chord.successor_changes_count	accumulator	The total number of successor changes. Frequent changes can identify a cluster problem.
chord.unstable_errors_count	accumulator	The total number of unstable-cluster errors returned to clients.
sync_with_master.elapsed_sec	accumulator	The total time, in seconds, spent synchronizing clusters.
sync_with_master.failures_count	accumulator	The total number of cluster synchronization failures.
sync_with_master.successes_count	accumulator	The total number of successful cluster synchronizations.

5.3.4. Environment

These metrics identify the node environment, license, and QuasarDB build.

Metric name	Type	Description	Notes
engine_build_date	label	The engine build date.
engine_version	label	The engine version.
hardware_concurrency_count	gauge	The number returned by std::thread::hardware_concurrency().
license.attribution_date_epoch	gauge	The license attribution time, in seconds after the Unix epoch.
license.expiration_date_epoch	gauge	The license expiration time, in seconds after the Unix epoch.
license.max_memory_bytes	gauge	The maximum number of memory bytes that the license permits for the node.
license.remaining_days_count	gauge	The number of days before the license expires.
license.support_until_epoch	gauge	The support expiration time, in seconds after the Unix epoch.
operating_system	label	A description of the operating system.
startup_epoch	gauge	The node start time, in seconds after the Unix epoch.

5.3.5. Indexes

These metrics describe use of the microindex subsystem during queries.

Metric name	Type	Description	Notes
queries.microindex.aggregation.match_count	accumulator	The total number of aggregations that used a microindex.
queries.microindex.aggregation.miss_count	accumulator	The total number of aggregations that could not use a microindex.
queries.microindex.filter.match_count	accumulator	The total number of filters, such as WHERE filters, that used a microindex.
queries.microindex.filter.miss_count	accumulator	The total number of filters, such as WHERE filters, that could not use a microindex.

5.3.6. Network

These metrics describe sessions, requests, and network traffic.

Metric name	Type	Description	Notes
network.current_users_count	gauge	The current number of users with an active session.
network.partitions_count	gauge	The number of network partitions.
network.sessions.available_count	gauge	The number of sessions that are available.
network.sessions.max_count	gauge	The maximum number of sessions.
network.sessions.unavailable_count	gauge	The number of sessions that are busy.
network.threads_per_partition_count	gauge	The number of threads in each network partition. By default, QuasarDB selects this value.
requests.failures_count	accumulator	The total number of failed requests.
requests.in_bytes	accumulator	The total number of request bytes received.
requests.out_bytes	accumulator	The total number of response bytes sent.
requests.slow_count	accumulator	The total number of requests that exceeded the configured slow-operation limit.
requests.successes_count	accumulator	The total number of successful requests.
requests.total_count	accumulator	The total number of received requests, including successful and failed requests.

5.3.7. Performance profiling

These metrics are available when network.profile_performance is enabled. Use them to compare the time spent in measured functions.

Metric name	Type	Description	Notes
perf.[name].[metric].total_ns	accumulator	The total time, in nanoseconds, for the specified metric in the specified function.
perf.[name].total_ns	accumulator	The total time, in nanoseconds, for the specified function.
perf.total_ns	accumulator	The total time, in nanoseconds, for all measured functions in the current performance trace.

5.3.8. Storage

These metrics describe the QuasarDB persistence layer.

Metric name	Type	Description	Notes
persistence.bucket_deletion_count	accumulator	The total number of delete operations on buckets.
persistence.bucket_insert_count	accumulator	The total number of insert operations on buckets.
persistence.bucket_not_found_count	accumulator	The total number of times that QuasarDB expected a bucket on disk but did not find it.
persistence.bucket_read_count	accumulator	The total number of read operations on buckets.
persistence.bucket_update_count	accumulator	The total number of update operations on buckets.
persistence.bucket.total_bytes	accumulator	The total number of bytes written for all buckets, including large buckets.
persistence.bucket.total_count	accumulator	The total number of bucket writes, including writes of large buckets.
persistence.bucket.total_us	accumulator	The total time, in microseconds, spent writing buckets, including large buckets.
persistence.cloud_local_cache_bytes	gauge	The current size, in bytes, of the local cache for RocksDB Cloud storage.
persistence.entries_count	gauge	The number of entries in the persistence layer. This value is related to, but usually greater than, the number of tables and buckets.
persistence.large_bucket.total_bytes	accumulator	The total number of bytes written for large buckets.
persistence.large_bucket.total_count	accumulator	The total number of large-bucket writes.
persistence.large_bucket.total_us	accumulator	The total time, in microseconds, spent writing large buckets.
persistence.persistent_cache_bytes	gauge	The current size, in bytes, of the persistent cache. This cache puts data from slow storage on faster storage. It is not the RocksDB Cloud local cache.
persistence.read_bytes	gauge	The total number of bytes that RocksDB read from storage.
persistence.root_path	label	The configured root path of the persistence layer.
persistence.ts_write.failures_count	accumulator	The total number of failed time-series write operations.
persistence.ts_write.successes_count	accumulator	The total number of successful time-series write operations.
persistence.utilized_bytes	gauge	The number of storage bytes used by RocksDB.
persistence.written_bytes	gauge	The total number of bytes that RocksDB wrote to storage.

5.3.9. Storage - Async pipelines

These metrics describe the async pipelines. In a metric name, [pipe_nb] identifies one pipeline, for example pipe_0.

Metric name	Type	Description	Notes
async_pipelines.busy_denied_count	accumulator	The number of async pushes that the server rejected because a pipeline was full.	The value is for the specified user.
async_pipelines.errors_count	accumulator	The number of async push errors that the server returned immediately.	The value is for the specified user.
async_pipelines.[pipe_nb].buffer_bytes	gauge	The current number of buffered bytes for the specified pipeline.
async_pipelines.[pipe_nb].buffer_count	gauge	The current number of buffers for the specified pipeline.
async_pipelines.[pipe_nb].buffer_rows_count	gauge	The current number of buffered rows for the specified pipeline.
async_pipelines.[pipe_nb].flush.latest_wait_ms	gauge	The interval, in milliseconds, before the most recent flush for the specified pipeline.
async_pipelines.[pipe_nb].flush.latest.buffer_bytes	gauge	The number of bytes in the most recent flush for the specified pipeline.
async_pipelines.[pipe_nb].flush.latest.buffer_count	gauge	The number of buffers in the most recent flush for the specified pipeline.
async_pipelines.[pipe_nb].flush.latest.buffer_rows_count	gauge	The number of rows in the most recent flush for the specified pipeline.
async_pipelines.[pipe_nb].flush.max_buffer_bytes	gauge	The largest number of bytes in one flush for the specified pipeline.
async_pipelines.[pipe_nb].flush.max_buffer_count	gauge	The largest number of buffers in one flush for the specified pipeline.
async_pipelines.[pipe_nb].flush.max_buffer_rows_count	gauge	The largest number of rows in one flush for the specified pipeline.
async_pipelines.[pipe_nb].flush.max_wait_ms	gauge	The longest interval, in milliseconds, before a flush for the specified pipeline.
async_pipelines.[pipe_nb].flush.total_buffer_bytes	accumulator	The total number of bytes flushed for the specified pipeline.
async_pipelines.[pipe_nb].flush.total_buffer_count	accumulator	The total number of buffers flushed for the specified pipeline.
async_pipelines.[pipe_nb].flush.total_buffer_rows_count	accumulator	The total number of rows flushed for the specified pipeline.
async_pipelines.[pipe_nb].flush.total_count	accumulator	The total number of flushes for the specified pipeline.
async_pipelines.[pipe_nb].flush.total_wait_ms	accumulator	The total time, in milliseconds, between flushes for the specified pipeline.
async_pipelines.[pipe_nb].max_buffer_bytes	gauge	The largest number of bytes held by the specified pipeline.
async_pipelines.[pipe_nb].max_buffer_count	gauge	The largest number of buffers held by the specified pipeline.
async_pipelines.[pipe_nb].max_buffer_rows_count	gauge	The largest number of rows held by the specified pipeline.
async_pipelines.[pipe_nb].write.failures.total_count	accumulator	The total number of write failures for the specified pipeline.
async_pipelines.[pipe_nb].write.latest_bytes	gauge	The number of bytes in the most recent write for the specified pipeline.
async_pipelines.[pipe_nb].write.latest_elapsed_us	gauge	The duration, in microseconds, of the most recent write for the specified pipeline.
async_pipelines.[pipe_nb].write.latest_rows_count	gauge	The number of rows in the most recent write for the specified pipeline.
async_pipelines.[pipe_nb].write.max_bytes	gauge	The largest number of bytes in one write for the specified pipeline.
async_pipelines.[pipe_nb].write.max_elapsed_us	gauge	The longest write duration, in microseconds, for the specified pipeline.
async_pipelines.[pipe_nb].write.max_rows_count	gauge	The largest number of rows in one write for the specified pipeline.
async_pipelines.[pipe_nb].write.successes.total_count	accumulator	The total number of successful writes for the specified pipeline.
async_pipelines.[pipe_nb].write.total_bytes	accumulator	The total number of bytes written for the specified pipeline.
async_pipelines.[pipe_nb].write.total_elapsed_us	accumulator	The total time, in microseconds, spent writing for the specified pipeline.
async_pipelines.[pipe_nb].write.total_rows_count	accumulator	The total number of rows written for the specified pipeline.
async_pipelines.flush.max_buffer_bytes	gauge	The largest number of bytes in one flush across all pipelines.
async_pipelines.flush.max_buffer_count	gauge	The largest number of buffers in one flush across all pipelines.
async_pipelines.flush.max_buffer_rows_count	gauge	The largest number of rows in one flush across all pipelines.
async_pipelines.flush.reason.deadline.total_count	accumulator	The total number of flushes started because the flush deadline was reached, across all pipelines.
async_pipelines.flush.reason.size.total_count	accumulator	The total number of flushes started because the buffered data reached half of the configured size limit, across all pipelines.
async_pipelines.flush.total_buffer_bytes	accumulator	The total number of bytes flushed across all pipelines.
async_pipelines.flush.total_buffer_count	accumulator	The total number of buffers flushed across all pipelines.
async_pipelines.flush.total_buffer_rows_count	accumulator	The total number of rows flushed across all pipelines.
async_pipelines.flush.total_count	accumulator	The total number of flushes across all pipelines.
async_pipelines.flush.total_wait_ms	accumulator	The total time, in milliseconds, between flushes across all pipelines.
async_pipelines.pull.max_bytes	gauge	The largest number of bytes pulled in one operation across all pipelines.
async_pipelines.pull.total_bytes	accumulator	The total number of bytes pulled across all pipelines.
async_pipelines.pull.total_count	accumulator	The total number of buffers pulled across all pipelines.
async_pipelines.push.max_bytes	gauge	The largest number of bytes pushed in one operation across all pipelines.
async_pipelines.push.total_bytes	accumulator	The total number of bytes pushed across all pipelines.
async_pipelines.push.total_count	accumulator	The total number of buffers pushed across all pipelines.
async_pipelines.write.failures.total_count	accumulator	The total number of write failures across all pipelines.
async_pipelines.write.max_bytes	gauge	The largest number of bytes in one write across all pipelines.
async_pipelines.write.max_elapsed_us	gauge	The longest write duration, in microseconds, across all pipelines.
async_pipelines.write.max_rows_count	gauge	The largest number of rows in one write across all pipelines.
async_pipelines.write.successes.total_count	accumulator	The total number of successful writes across all pipelines.
async_pipelines.write.total_bytes	accumulator	The total number of bytes written across all pipelines.
async_pipelines.write.total_elapsed_us	accumulator	The total time, in microseconds, spent writing across all pipelines.
async_pipelines.write.total_rows_count	accumulator	The total number of rows written across all pipelines.

5.3.10. Storage - Backups

These metrics describe storage backups.

Metric name	Type	Description	Notes
backup.elapsed_sec	accumulator	The total time, in seconds, spent on backups.
backup.failures_count	accumulator	The total number of backup failures.
backup.successes_count	accumulator	The total number of successful background backups.
backup.total_bytes	accumulator	The total number of bytes written during backups.

5.3.11. Storage - Optimization

These metrics describe storage operations that maintain performance and manage the data lifecycle.

Metric name	Type	Description	Notes
compact.cancelations_count	accumulator	The total number of canceled background compactions.
compact.elapsed_sec	accumulator	The total time, in seconds, spent on compaction.
compact.failures_count	accumulator	The total number of background compaction failures.
compact.successes_count	accumulator	The total number of successful, manually started background compactions.
trim.cancelations_count	accumulator	The total number of canceled background trim operations.
trim.elapsed_sec	accumulator	The total time, in seconds, spent on background trim operations.
trim.failures_count	accumulator	The total number of background trim failures.
trim.successes_count	accumulator	The total number of successful background trim operations.

5.3.12. Storage - RocksDB

These metrics expose low-level RocksDB properties and counters.

Metric name	Type	Description	Notes
persistence.rocksdb.actual_delayed_write_rate	gauge	The current delayed-write rate, in bytes per second. A value of 0 means that writes are not delayed.
persistence.rocksdb.block_cache_capacity_bytes	gauge	The block-cache capacity, in bytes.
persistence.rocksdb.block_cache_pinned_usage_bytes	gauge	The number of block-cache bytes used by pinned entries.
persistence.rocksdb.block_cache_usage_bytes	gauge	The number of bytes used by entries in the block cache.
persistence.rocksdb.compact_read_bytes	accumulator	The total number of bytes read during compaction.
persistence.rocksdb.compact_read_marked_bytes	accumulator	The total number of bytes read during compactions with the manually-marked reason.
persistence.rocksdb.compact_read_periodic_bytes	accumulator	The total number of bytes read during periodic compactions.
persistence.rocksdb.compact_read_ttl_bytes	accumulator	The total number of bytes read during TTL compactions.
persistence.rocksdb.compact_write_bytes	accumulator	The total number of bytes written during compaction.
persistence.rocksdb.compact_write_marked_bytes	accumulator	The total number of bytes written during compactions with the manually-marked reason.
persistence.rocksdb.compact_write_periodic_bytes	accumulator	The total number of bytes written during periodic compactions.
persistence.rocksdb.compact_write_ttl_bytes	accumulator	The total number of bytes written during TTL compactions.
persistence.rocksdb.compaction_pending_count	gauge	A value of 1 if a compaction is pending. Otherwise, the value is 0.
persistence.rocksdb.cur_size_active_memtable_bytes	gauge	The approximate size, in bytes, of the active memtable.
persistence.rocksdb.cur_size_all_memtables_bytes	gauge	The approximate size, in bytes, of the active memtable and unflushed immutable memtables.
persistence.rocksdb.current_super_version_number	gauge	The current LSM-tree version number. The value increases after an LSM-tree change and resets to 0 after a database restart.
persistence.rocksdb.estimate_live_data_size_bytes	gauge	The estimated size, in bytes, of live data. For BlobDB, this value includes live data in the current blob files.
persistence.rocksdb.estimate_num_keys_count	gauge	The estimated number of keys in the active memtable, unflushed immutable memtables, and storage.
persistence.rocksdb.estimate_pending_compaction_bytes	gauge	The estimated number of bytes that level-based compaction must rewrite to put all levels below their target sizes.
persistence.rocksdb.estimate_table_readers_mem_bytes	gauge	The estimated number of bytes used to read SST tables. This value excludes block-cache memory.
persistence.rocksdb.flush_write_bytes	accumulator	The total number of bytes written during memtable flushes.
persistence.rocksdb.get_hit_l0_count	accumulator	The total number of Get() operations served by L0.
persistence.rocksdb.get_hit_l1_count	accumulator	The total number of Get() operations served by L1.
persistence.rocksdb.get_hit_l2_and_up_count	accumulator	The total number of Get() operations served by L2 or a higher level.
persistence.rocksdb.last_level_read_bytes	accumulator	The total number of bytes read from the last LSM-tree level.
persistence.rocksdb.last_level_read_count	accumulator	The total number of read operations on the last LSM-tree level.
persistence.rocksdb.live_blob_file_garbage_size_bytes	gauge	The number of garbage bytes in blob files in the current version.
persistence.rocksdb.live_blob_file_size_bytes	gauge	The total size, in bytes, of blob files in the current version.
persistence.rocksdb.live_non_bottommost_sst_files_size_bytes	gauge	The total size, in bytes, of SST files above the bottom level in the current LSM tree.
persistence.rocksdb.live_sst_files_size_bytes	gauge	The total size, in bytes, of SST files in the current column-family version.
persistence.rocksdb.memtable_flush_pending_count	gauge	A value of 1 if a memtable flush is pending. Otherwise, the value is 0.
persistence.rocksdb.min_log_number_to_keep	gauge	The minimum log-file number that RocksDB must keep.
persistence.rocksdb.min_obsolete_sst_number_to_keep	gauge	The minimum obsolete SST-file number that RocksDB must keep. The maximum uint64_t value means that RocksDB can delete all obsolete SST files.
persistence.rocksdb.non_last_level_read_bytes	accumulator	The total number of bytes read from LSM-tree levels above the last level.
persistence.rocksdb.non_last_level_read_count	accumulator	The total number of read operations on LSM-tree levels above the last level.
persistence.rocksdb.num_blob_files_count	gauge	The number of blob files in the current version.
persistence.rocksdb.num_deletes_active_memtable_count	gauge	The number of delete entries in the active memtable.
persistence.rocksdb.num_deletes_imm_memtables_count	gauge	The number of delete entries in unflushed immutable memtables.
persistence.rocksdb.num_entries_active_memtable_count	gauge	The number of entries in the active memtable.
persistence.rocksdb.num_entries_imm_memtables_count	gauge	The number of entries in unflushed immutable memtables.
persistence.rocksdb.num_immutable_mem_table_count	gauge	The number of immutable memtables that are not flushed.
persistence.rocksdb.num_immutable_mem_table_flushed_count	gauge	The number of immutable memtables that are flushed.
persistence.rocksdb.num_live_versions_count	gauge	The number of live RocksDB versions. A large value can identify iterators or unfinished compactions that prevent deletion of SST files.
persistence.rocksdb.num_running_compactions_count	gauge	The number of compactions in progress.
persistence.rocksdb.num_running_flushes_count	gauge	The number of flushes in progress.
persistence.rocksdb.num_snapshots_count	accumulator	The number of unreleased database snapshots.
persistence.rocksdb.obsolete_sst_files_size_bytes	gauge	The total size, in bytes, of obsolete SST files that RocksDB did not delete or schedule for deletion.
persistence.rocksdb.oldest_snapshot_sequence	gauge	The sequence number of the oldest unreleased snapshot.
persistence.rocksdb.oldest_snapshot_time_epoch	gauge	The creation time of the oldest unreleased snapshot, in seconds after the Unix epoch.
persistence.rocksdb.persistent_cache_hit_count	accumulator	The total number of persistent-cache hits.
persistence.rocksdb.persistent_cache_miss_count	accumulator	The total number of persistent-cache misses.
persistence.rocksdb.size_all_memtables_bytes	gauge	The approximate size, in bytes, of the active memtable and unflushed immutable memtables.
persistence.rocksdb.total_blob_file_size_bytes	gauge	The total size, in bytes, of blob files in all versions.
persistence.rocksdb.total_sst_files_size_bytes	gauge	The total size, in bytes, of SST files in all column-family versions. Collection of this value can slow queries when the database has many files.
rocksdb.background_errors_count	accumulator	The total number of RocksDB background errors.

5.3.13. Metric name suffixes

A metric-name suffix identifies the unit or value type.

Suffix	Meaning
_ns	A duration in nanoseconds.
_us	A duration in microseconds.
_ms	A duration in milliseconds.
_sec	A duration in seconds.
_epoch	A time in seconds after the Unix epoch.
_bytes	A number of bytes.
_count	A number of items, operations, or events.
_total or total_	A value accumulated during the life of the process.