Only Application metrics are listed here. These are kept compatible
across releases. Developer metrics are hidden unless
search.info-developer-visible is set to yes,
and may change or be removed in any release without notice.
Some metrics are reported only when they apply. The
coordinator_* metrics appear only when the module is loaded
with --use-coordinator. Metrics ending in
_latency_usec additionally appear only after at least one
sample has been recorded.
Valkey prefixes each field with search_ and each section
with search_ in INFO output, so
compatibility_usage appears as
search_compatibility_usage in section
search_compatibility.
| Metric Name | Section | Unit | Description |
|---|---|---|---|
compatibility_usage |
compatibility |
String | Comma-separated label=count
list of legacy behaviors used because
search.emulate-release is below the release that fixed them
(see below) |
coordinator_bytes_in |
coordinator |
Bytes | Total size (in bytes) of incoming gRPC response messages received by the coordinator client from remote nodes |
coordinator_bytes_out |
coordinator |
Bytes | Total size (in bytes) of outgoing gRPC request messages sent by the coordinator client to remote nodes |
coordinator_client_get_global_metadata_failure_count |
coordinator |
Count | Count of failed client requests to get global metadata |
coordinator_client_get_global_metadata_failure_latency_usec |
coordinator |
Microseconds | Latency distribution (in microseconds) for failed client metadata requests |
coordinator_client_get_global_metadata_success_count |
coordinator |
Count | Count of successful client requests to get global metadata |
coordinator_client_get_global_metadata_success_latency_usec |
coordinator |
Microseconds | Latency distribution (in microseconds) for successful client metadata requests |
coordinator_client_search_index_partition_failure_count |
coordinator |
Count | Count of failed client searches on index partitions |
coordinator_client_search_index_partition_failure_latency_usec |
coordinator |
Microseconds | Latency distribution (in microseconds) for failed partition searches |
coordinator_client_search_index_partition_success_count |
coordinator |
Count | Count of successful client searches on index partitions |
coordinator_client_search_index_partition_success_latency_usec |
coordinator |
Microseconds | Latency distribution (in microseconds) for successful partition searches |
coordinator_server_get_global_metadata_failure_count |
coordinator |
Count | Count of failed server requests to get global metadata |
coordinator_server_get_global_metadata_failure_latency_usec |
coordinator |
Microseconds | Latency distribution (in microseconds) for failed server metadata requests |
coordinator_server_get_global_metadata_success_count |
coordinator |
Count | Count of successful server requests to get global metadata |
coordinator_server_get_global_metadata_success_latency_usec |
coordinator |
Microseconds | Latency distribution (in microseconds) for successful server metadata requests |
coordinator_server_listening_port |
coordinator |
Port | The port number the coordinator server is listening on |
coordinator_server_search_index_partition_cpu_time_usec |
coordinator |
Microseconds | Cumulative CPU time (in microseconds) spent handling remote query operations on this node |
coordinator_server_search_index_partition_failure_count |
coordinator |
Count | Count of failed server searches on index partitions |
coordinator_server_search_index_partition_failure_latency_usec |
coordinator |
Microseconds | Latency distribution (in microseconds) for failed server partition searches |
coordinator_server_search_index_partition_success_count |
coordinator |
Count | Count of successful server searches on index partitions |
coordinator_server_search_index_partition_success_latency_usec |
coordinator |
Microseconds | Latency distribution (in microseconds) for successful server partition searches |
coordinator_threads_cpu_time_sec |
coordinator |
Seconds | Cumulative CPU time consumed by coordinator (gRPC) threads |
hnsw_add_exceptions_count |
hnswlib |
Count | Count of exceptions during HNSW vector additions |
hnsw_create_exceptions_count |
hnswlib |
Count | Count of exceptions during HNSW vector creation |
hnsw_modify_exceptions_count |
hnswlib |
Count | Count of exceptions during HNSW vector modifications |
hnsw_remove_exceptions_count |
hnswlib |
Count | Count of exceptions during HNSW vector removals |
hnsw_search_exceptions_count |
hnswlib |
Count | Count of exceptions during HNSW vector searches |
number_of_attributes |
index_stats |
Count | Total count of all attributes across all search indexes |
number_of_indexes |
index_stats |
Count | Total number of search indexes |
num_cursors |
cursors |
Count | Number of open cursors created by
WITHCURSOR |
total_active_write_threads |
index_stats |
Count | Number of active writer threads (0 if suspended, otherwise the writer thread pool size) |
total_indexed_documents |
index_stats |
Count | Total number of indexed documents across all search indexes |
background_indexing_status |
indexing |
String | Background indexing status: IN_PROGRESS or NO_ACTIVITY |
flat_vector_index_search_latency_usec |
latency |
Microseconds | Latency distribution (in microseconds) for flat vector index searches |
hnsw_vector_index_search_latency_usec |
latency |
Microseconds | Latency distribution (in microseconds) for HNSW vector index searches |
index_reclaimable_memory |
memory |
Bytes | Track memory that can be reclaimed after vector deletions |
used_memory_bytes |
memory |
Bytes | Total memory used by the module (in bytes) |
used_memory_human |
memory |
String | Total memory used by the module (in human readable format) |
successful_requests_count |
query |
Count | Total count of successful query requests |
failure_requests_count |
query |
Count | Total count of failed query requests |
hybrid_requests_count |
query |
Count | Count of KNN queries that also have a
non-vector filter, for example
@color:{red}=>[KNN 10 @vec $v], including an
FT.HYBRID VSIM arm with FILTER.
It does not count FT.HYBRID commands as such |
nonvector_requests_count |
query |
Count | Number of requests that are exclusively
non-vector search queries, including vector range
(VECTOR_RANGE) queries and the SEARCH arm of
FT.HYBRID |
vector_requests_count |
query |
Count | Number of KNN queries without a filter,
for example *=>[KNN 10 @vec $v]. An
FT.HYBRID VSIM arm without FILTER
is not counted |
inline_filtering_requests_count |
query |
Count | Count of queries using inline filtering |
prefiltering_requests_count |
query |
Count | Count of queries using pre-filtering |
result_record_dropped_count |
query |
Count | Tracks records dropped when FT.SEARCH results exceed configured limits |
vector_range_results_fetched_limited_count |
query |
Count | Count of VECTOR_RANGE queries
whose results were truncated on this shard. See Result limit. |
rdb_load_failure_cnt |
rdb |
Count | Number of failed RDB load operations |
rdb_load_success_cnt |
rdb |
Count | Number of successful RDB load operations |
rdb_save_failure_cnt |
rdb |
Count | Number of failed RDB save operations |
rdb_save_success_cnt |
rdb |
Count | Number of successful RDB save operations |
string_interning_store_size |
string_interning |
Count | Number of unique strings in the interning store |
query_queue_size |
thread-pool |
Count | Current size of the reader thread pool queue |
reader_resumed_cnt |
thread-pool |
Count | Number of times the reader worker thread pool was resumed |
used_read_cpu |
thread-pool |
Percent | Average CPU percentage used by reader threads (returns -1 if unavailable) |
used_write_cpu |
thread-pool |
Percent | Average CPU percentage used by writer threads (returns -1 if unavailable) |
worker_pool_suspend_cnt |
thread-pool |
Count | Number of times the worker thread pool was suspended |
writer_queue_size |
thread-pool |
Count | Current size of the writer thread pool queue |
writer_resumed_cnt |
thread-pool |
Count | Number of times the writer worker thread pool was resumed |
writer_suspension_expired_cnt |
thread-pool |
Count | Number of times writer suspension expired due to timeout |
vector_registry_dedup_cnt |
vector_registry |
Count | Count of vector updates that reused an already registered copy of an identical vector |
vector_registry_entry_cnt |
vector_registry |
Count | Number of vectors currently tracked by the vector registry |
vector_registry_pending_unshare_cnt |
vector_registry |
Count | Number of vectors queued to stop being shared with the engine, after their index was dropped |
vector_registry_shared_externally_cnt |
vector_registry |
Count | Count of vectors successfully shared with the engine’s hash field, so only one copy is kept in memory |
vector_registry_shared_externally_errors |
vector_registry |
Count | Count of failed attempts to share a vector with the engine’s hash field |
vector_registry_sharing_active |
vector_registry |
Boolean | 1 if vector memory sharing with the engine is enabled, otherwise 0 |
The compatibility_usage metric is a comma-separated list
of label=count pairs. Each count is how often a legacy,
incompatible behavior was used because
search.emulate-release is set below the release that fixed
it. See Compatibility
Defects for the full list of labels and what each one changes. The
metric is always reported; a label appears in the list only after its
legacy behavior is first used, and an absent label has a count of 0.