Metrics

About 9 min

Metrics

EventStoreDB collects metrics in Prometheus format, available on the /metrics endpoint. Prometheus can be configured to scrape this endpoint directly. The metrics are configured in metricsconfig.json.

In addition, EventStoreDB can actively export metrics to a specified endpoint using the OpenTelemetry Protocol (OTLP). Commercial

Metrics reference

Caches

Cache hits and misses

EventStoreDB tracks cache hits/misses metrics for stream-info and chunk caches.

Time series	Type	Description
`eventstore_cache_hits_misses{cache=<CACHE_NAME>,kind=<"hits"\|"misses">}`	Counter	Total hits/misses on CACHE_NAME cache

Example configuration:

"CacheHitsMisses": {
  "StreamInfo": true,
  "Chunk": false
}

Example output:

# TYPE eventstore_cache_hits_misses counter
eventstore_cache_hits_misses{cache="stream-info",kind="hits"} 104329 1688157489545
eventstore_cache_hits_misses{cache="stream-info",kind="misses"} 117 1688157489545

Dynamic cache resources

Certain caches that EventStoreDB uses are dynamic in nature i.e. their capacity scales up/down during their lifetime. EventStoreDB records metrics for resources being used by each such dynamic cache.

Time series	Type	Description
`eventstore_cache_resources_bytes{cache=<CACHE_NAME>,kind=<"capacity"\|"size">}`	Gauge	Current capacity/size of CACHE_NAME cache in bytes
`eventstore_cache_resources_entries{cache=<CACHE_NAME>,kind="count"}`	Gauge	Current number of entries in CACHE_NAME cache

Example configuration:

"CacheResources": true

Example output:

# TYPE eventstore_cache_resources_bytes gauge
# UNIT eventstore_cache_resources_bytes bytes
eventstore_cache_resources_bytes{cache="LastEventNumber",kind="capacity"} 50000000 1688157491029
eventstore_cache_resources_bytes{cache="LastEventNumber",kind="size"} 15804 1688157491029

# TYPE eventstore_cache_resources_entries gauge
# UNIT eventstore_cache_resources_entries entries
eventstore_cache_resources_entries{cache="LastEventNumber",kind="count"} 75 1688157491029

Checkpoints

Time series	Type	Description
`eventstore_checkpoints{name=<CHECKPOINT_NAME>,read="non-flushed"}`	Gauge	Value for CHECKPOINT_NAME checkpoint

Example configuration:

"Checkpoints": {
  "Replication": true,
  "Chaser": false,
  "Epoch": false,
  "Index": false,
  "Proposal": false,
  "Truncate": false,
  "Writer": false,
  "StreamExistenceFilter": false
}

Example output:

# TYPE eventstore_checkpoints gauge
eventstore_checkpoints{name="replication",read="non-flushed"} 613363 1688054162478

Elections Count

This metric tracks the number of elections that have been completed.

Time series	Type	Description
`eventstore_elections_count`	Counter	Elections count in a cluster

Example configuration:

"ElectionsCount": true

Example output:

# TYPE eventstore_elections_count counter
eventstore_elections_count 0 1710188996949

Events

These metrics track events written to and read from the server, including reads from caches.

Time series	Type	Description
`eventstore_io_bytes{activity="read"}`	Counter	Event bytes read
`eventstore_io_events{activity=<"read"\|"written">}`	Counter	Events read/written

Example configuration:

"Events": {
  "Read": false,
  "Written": true
}

Example output:

# TYPE eventstore_io_events counter
# UNIT eventstore_io_events events
eventstore_io_events{activity="written"} 320 1687963622074

Gossip

Measures the round trip latency and processing time of gossip. Usually a node pushes new gossip to other nodes periodically or when its view of the cluster changes. Sometimes nodes pull gossip from each other if there is a suspected network problem.

Gossip latency

Time series	Type	Description
`eventstore_gossip_latency_seconds_bucket{activity="pull-from-peer",status=<"successful"\|"failed">,le=<DURATION>}`	Histogram	Number of gossips pulled from peers with latency less than or equal to DURATION in seconds
`eventstore_gossip_latency_seconds_bucket{activity="push-to-peer",status=<"successful"\|"failed">,le=<DURATION>}`	Histogram	Number of gossips pushed to peers with latency less than or equal to DURATION in seconds

Gossip processing

Time Series	Type	Description
`eventstore_gossip_processing_duration_seconds_bucket{` `activity="push-from-peer",` `status=<"successful"\|"failed">,` `le=<DURATION>}`	Histogram	Number of gossips pushed from peers that took less than or equal to DURATION in seconds to process
`eventstore_gossip_processing_duration_seconds_bucket{` `activity="request-from-peer",` `status=<"successful"\|"failed">,` `le=<DURATION>}`	Histogram	Number of gossip requests from peers that took less than or equal to DURATION in seconds to process
`eventstore_gossip_processing_duration_seconds_bucket{` `activity="request-from-grpc-client",` `status=<"successful"\|"failed">,` `le=<DURATION>}`	Histogram	Number of gossip requests from gRPC clients that took less than or equal to DURATION in seconds to process
`eventstore_gossip_processing_duration_seconds_bucket{` `activity="request-from-http-client",` `status=<"successful"\|"failed">,` `le=<DURATION>}`	Histogram	Number of gossip requests from HTTP clients that took less than or equal to DURATION in seconds to process

Example configuration:

"Gossip": {
  "PullFromPeer": false,
  "PushToPeer": true,
  "ProcessingPushFromPeer": false,
  "ProcessingRequestFromPeer": false,
  "ProcessingRequestFromGrpcClient": false,
  "ProcessingRequestFromHttpClient": false
}

Example output:

# TYPE eventstore_gossip_latency_seconds histogram
# UNIT eventstore_gossip_latency_seconds seconds
eventstore_gossip_latency_seconds_bucket{activity="push-to-peer",status="successful",le="0.005"} 8 1687972306948

Incoming gRPC calls

Time series	Type	Description
`eventstore_current_incoming_grpc_calls`	Gauge	Inflight gRPC calls i.e. gRPC requests that have started on the server but not yet stopped
`eventstore_incoming_grpc_calls{kind="total"}`	Counter	Total gRPC requests served
`eventstore_incoming_grpc_calls{kind="failed"}`	Counter	Total gRPC requests failed
`eventstore_incoming_grpc_calls{` `kind="unimplemented"}`	Counter	Total gRPC requests made to unimplemented methods
`eventstore_incoming_grpc_calls{` `kind="deadline-exceeded"}`	Counter	Total gRPC requests for which deadline have exceeded

Example configuration:

"IncomingGrpcCalls": {
  "Current": true,
  "Total": false,
  "Failed": true,
  "Unimplemented": false,
  "DeadlineExceeded": false
}

Example output:

# TYPE eventstore_current_incoming_grpc_calls gauge
eventstore_current_incoming_grpc_calls 1 1687963622074

# TYPE eventstore_incoming_grpc_calls counter
eventstore_incoming_grpc_calls{kind="failed"} 1 1687962877623

Client protocol gRPC methods

In addition, EventStoreDB also records metrics for each of client protocol gRPC methods: StreamRead, StreamAppend, StreamBatchAppend, StreamDelete and StreamTombstone. They are grouped together according to the mapping defined in the configuration.

Time series	Type	Description
`eventstore_grpc_method_duration_seconds_bucket{` `activity=<LABEL>,` `status="successful"\|"failed",` `le=<DURATION>}`	Histogram	Number of LABEL gRPC requests that took less than or equal to DURATION in seconds to process

Example configuration:

"GrpcMethods": {
  "StreamAppend": "append",
  "StreamBatchAppend": "append",
  // leaving label as blank will disable metric collection
  "StreamRead": "",
  "StreamDelete": "",
  "StreamTombstone": ""
}

Example output:

# TYPE eventstore_grpc_method_duration_seconds histogram
# UNIT eventstore_grpc_method_duration_seconds seconds
eventstore_grpc_method_duration_seconds_bucket{activity="append",status="successful",le="1E-06"} 0 1688157491029
eventstore_grpc_method_duration_seconds_bucket{activity="append",status="successful",le="1E-05"} 0 1688157491029
eventstore_grpc_method_duration_seconds_bucket{activity="append",status="successful",le="0.0001"} 129 1688157491029
eventstore_grpc_method_duration_seconds_bucket{activity="append",status="successful",le="0.001"} 143 1688157491029
eventstore_grpc_method_duration_seconds_bucket{activity="append",status="successful",le="0.01"} 168 1688157491029
eventstore_grpc_method_duration_seconds_bucket{activity="append",status="successful",le="0.1"} 169 1688157491029
eventstore_grpc_method_duration_seconds_bucket{activity="append",status="successful",le="1"} 169 1688157491029
eventstore_grpc_method_duration_seconds_bucket{activity="append",status="successful",le="10"} 169 1688157491029
eventstore_grpc_method_duration_seconds_bucket{activity="append",status="successful",le="+Inf"} 169 1688157491029

Kestrel

Time series	Type	Description
`eventstore_kestrel_connections`	Gauge	Number of open kestrel connections

Example configuration:

"Kestrel": {
  "ConnectionCount": true
}

Example output:

# TYPE eventstore_kestrel_connections gauge
eventstore_kestrel_connections 1 1688070655500

Persistent Subscriptions

Persistent subscription metrics track the statistics for the persistent subscriptions.

Time series	Type	Description
`eventstore_persistent_sub_connections` `{event_stream_id=<STREAM_NAME>,group_name=<GROUP_NAME>}`	Gauge	Number of connections
`eventstore_persistent_sub_parked_messages` `{event_stream_id=<STREAM_NAME>,group_name=<GROUP_NAME>}`	Gauge	Number of parked messages
`eventstore_persistent_sub_in_flight_messages` `{event_stream_id=<STREAM_NAME>,group_name=<GROUP_NAME>}`	Gauge	Number of messages in flight
`eventstore_persistent_sub_oldest_parked_message_seconds` `{event_stream_id=<STREAM_NAME>,group_name=<GROUP_NAME>}`	Gauge	Oldest parked message age in seconds
`eventstore_persistent_sub_items_processed` `{event_stream_id=<STREAM_NAME>,group_name=<GROUP_NAME>}`	Counter	Total items processed
`eventstore_persistent_sub_last_known_event_number` `{event_stream_id=<STREAM_NAME>,group_name=<GROUP_NAME>}`	Counter	Last known event number (streams other than `$all`)
`eventstore_persistent_sub_last_known_event_commit_position` `{event_stream_id=<STREAM_NAME>,group_name=<GROUP_NAME>}`	Counter	Last known event's commit position (`$all` stream only)
`eventstore_persistent_sub_checkpointed_event_number` `{event_stream_id=<STREAM_NAME>,group_name=<GROUP_NAME>}`	Counter	Last checkpointed event number (streams other than `$all)`
`eventstore_persistent_sub_checkpointed_event_commit_position` `{event_stream_id=<STREAM_NAME>,group_name=<GROUP_NAME>}`	Counter	Last checkpointed event's commit position (`$all` stream only)

Example configuration:

"PersistentSubscriptionStats": true,

Example output:

# TYPE eventstore_persistent_sub_connections gauge
eventstore_persistent_sub_connections{event_stream_id="test-stream",group_name="group1"} 1 1720172078179
eventstore_persistent_sub_connections{event_stream_id="$all",group_name="group1"} 1 1720172078179

# TYPE eventstore_persistent_sub_parked_messages gauge
eventstore_persistent_sub_parked_messages{event_stream_id="test-stream",group_name="group1"} 0 1720172078179
eventstore_persistent_sub_parked_messages{event_stream_id="$all",group_name="group1"} 0 1720172078179

# TYPE eventstore_persistent_sub_in_flight_messages gauge
eventstore_persistent_sub_in_flight_messages{event_stream_id="test-stream",group_name="group1"} 0 1720172078179
eventstore_persistent_sub_in_flight_messages{event_stream_id="$all",group_name="group1"} 0 1720172078179

# TYPE eventstore_persistent_sub_oldest_parked_message_seconds gauge
eventstore_persistent_sub_oldest_parked_message_seconds{event_stream_id="test-stream",group_name="group1"} 0 1720172078179
eventstore_persistent_sub_oldest_parked_message_seconds{event_stream_id="$all",group_name="group1"} 0 1720172078179

# TYPE eventstore_persistent_sub_items_processed counter
eventstore_persistent_sub_items_processed{event_stream_id="test-stream",group_name="group1"} 0 1720172078179
eventstore_persistent_sub_items_processed{event_stream_id="$all",group_name="group1"} 5 1720172078179

# TYPE eventstore_persistent_sub_last_known_event_number counter
eventstore_persistent_sub_last_known_event_number{event_stream_id="test-stream",group_name="group1"} 0 1720172078179

# TYPE eventstore_persistent_sub_last_known_event_commit_position counter
eventstore_persistent_sub_last_known_event_commit_position{event_stream_id="$all",group_name="group1"} 4113 1720172078179

# TYPE eventstore_persistent_sub_checkpointed_event_number counter
eventstore_persistent_sub_checkpointed_event_number{event_stream_id="test-stream",group_name="group1"} 0 1720172078179

# TYPE eventstore_persistent_sub_checkpointed_event_commit_position counter
eventstore_persistent_sub_checkpointed_event_commit_position{event_stream_id="$all",group_name="group1"} 4013 1720172078179

Process

EventStoreDB collects key metrics about the running process.

Time Series	Type	Description
`eventstore_proc_up_time{pid=<PID>}`	Counter	Time in seconds this process has been running for. PID is process Id of EventStoreDB process
`eventstore_proc_cpu`	Gauge	Process CPU usage
`eventstore_proc_thread_count`	Gauge	Current number of threadpool threads (ThreadPool.ThreadCount)
`eventstore_proc_thread_pool_pending_work_item_count`	Gauge	Current number of items that are queued to be processed by threadpool threads (ThreadPool.PendingWorkItemCount)
`eventstore_proc_contention_count`	Counter	Total number of times there was contention when trying to take monitor's lock (Monitor.LockContentionCount)
`eventstore_proc_exception_count`	Counter	Total number of exceptions thrown
`eventstore_gc_time_in_gc`	Gauge	Percentage of CPU time spent collecting garbage during last garbage collection
`eventstore_gc_heap_size_bytes`	Gauge	Heap size in bytes
`eventstore_gc_heap_fragmentation`	Gauge	Percentage of heap fragmentation during last garbage collection
`eventstore_gc_total_allocated`	Counter	Total allocated bytes over the lifetime of this process
`eventstore_gc_pause_duration_max_seconds{` `range=<RANGE>}`	RecentMax	Recent maximum garbage collection pause in seconds. This measures the times that the execution engine is paused for GC
`eventstore_gc_generation_size_bytes{` `generation=<"gen0"\|"gen1"\|"gen2"\|"loh">}`	Gauge	Size of each generation in bytes
`eventstore_gc_collection_count{` `generation=<"gen0"\|"gen1"\|"gen2">}`	Counter	Number of garbage collections from each generation
`eventstore_proc_mem_bytes{` `kind=<"working-set"\|"paged-bytes"\|"virtual-bytes">}`	Gauge	Size in bytes of the working set, paged or virtual memory
`eventstore_disk_io_bytes{` `activity=<"read"\|"written">}`	Counter	Number of bytes read from/written to the disk
`eventstore_disk_io_operations{` `activity=<"read"\|"written">}`	Counter	Number of OS read/write operations issued to the disk

Example configuration:

"Process": {
  "UpTime": false,
  "Cpu": false,
  "MemWorkingSet": false,
  "MemPagedBytes": false,
  "MemVirtualBytes": false,
  "ThreadCount": true,
  "ThreadPoolPendingWorkItemCount": false,
  "LockContentionCount": true,
  "ExceptionCount": false,
  "Gen0CollectionCount": false,
  "Gen1CollectionCount": false,
  "Gen2CollectionCount": false,
  "Gen0Size": false,
  "Gen1Size": false,
  "Gen2Size": false,
  "LohSize": false,
  "TimeInGc": false,
  "GcPauseDuration": true,
  "HeapSize": false,
  "HeapFragmentation": false,
  "TotalAllocatedBytes": false,
  "DiskReadBytes": false,
  "DiskReadOps": false,
  "DiskWrittenBytes": false,
  "DiskWrittenOps": false
}

Example output:

# TYPE eventstore_proc_thread_count gauge
eventstore_proc_thread_count 15 1688070655500

# TYPE eventstore_proc_contention_count counter
eventstore_proc_contention_count 297 1688147136862

# TYPE eventstore_gc_pause_duration_max_seconds gauge
# UNIT eventstore_gc_pause_duration_max_seconds seconds
eventstore_gc_pause_duration_max_seconds{range="16-20 seconds"} 0.0485873 1688147136862

Projections

Projection metrics track the statistics for projections.

Time series	Type	Description
`eventstore_projection_events_processed_after_restart_total` `{projection=<PROJECTION_NAME>}`	Counter	Projection event processed count after restart
`eventstore_projection_progress{projection=<PROJECTION_NAME>}`	Gauge	Projection progress 0 - 1, where 1 = projection progress at 100%
`eventstore_projection_running{projection=<PROJECTION_NAME>}`	Gauge	If 1, projection is in 'Running' state
`eventstore_projection_status{projection=<PROJECTION_NAME>,status=<PROJECTION_STATUS>}`	Gauge	If 1, projection is in specified state

Status can have one of the following statuses:

Running
Faulted
Stopped

Example configuration:

"ProjectionStats": true

Example output:

# TYPE eventstore_projection_events_processed_after_restart_total counter
eventstore_projection_events_processed_after_restart_total{projection="$by_category"} 83 1719526306309

# TYPE eventstore_projection_progress gauge
eventstore_projection_progress{projection="$stream_by_category"} 1 1719526306309

# TYPE eventstore_projection_running gauge
eventstore_projection_running{projection="$by_category"} 1 1719526306309

# TYPE eventstore_projection_status gauge
eventstore_projection_status{projection="$by_category",status="Running"} 1 1719526306309
eventstore_projection_status{projection="$by_category",status="Faulted"} 0 1719526306309
eventstore_projection_status{projection="$by_category",status="Stopped"} 0 1719526306309

Queues

EventStoreDB uses various queues for asynchronous processing for which it also collects different metrics. In addition, EventStoreDB allows users to group queues and monitor them as a unit.

Time series	Type	Description
`eventstore_queue_busy_seconds{queue=<QUEUE_GROUP>}`	Counter	Total time spent processing in seconds, averaged across the queues in the QUEUE_GROUP. The rate of this metric is therefore the average busyness of the group during the period (from 0-1 s/s)
`eventstore_queue_queueing_duration_max_seconds{` `name=<QUEUE_GROUP>,range=<RANGE>}`	RecentMax	Recent maximum time in seconds for which any item was queued in queues belonging to the QUEUE_GROUP. This is essentially the length of the longest queue in the group in seconds
`eventstore_queue_processing_duration_seconds_bucket{` `message_type=<TYPE>,queue=<QUEUE_GROUP>,le=<DURATION>}`	Histogram	Number of messages of type TYPE processed by QUEUE_GROUP group that took less than or equal to DURATION in seconds

QueueLabels setting within metricsconfig.json can be used to group queues, based on regex which gets matched on queue names, and label them for metrics reporting. Capture groups are also supported. Message types can be grouped in the same way in the MessageTypes setting in metricsconfig.json.

Note

Enabling Queues.Processing can cause a lot more time series to be generated, according to the QueueLabels and MessageTypes configuration.

Example configuration:

"Queues": {
  "Busy": true,
  "Length": true,
  "Processing": false
}

"QueueLabels": [
  {
    "Regex": "StorageReaderQueue #.*",
    "Label": "Readers"
  },
  {
    "Regex": ".*",
    "Label": "Others"
  }
]

Example output:

# TYPE eventstore_queue_busy_seconds counter
# UNIT eventstore_queue_busy_seconds seconds
eventstore_queue_busy_seconds{queue="Readers"} 1.04568158125 1688157491029
eventstore_queue_busy_seconds{queue="Others"} 0 1688157491029

# TYPE eventstore_queue_queueing_duration_max_seconds gauge
# UNIT eventstore_queue_queueing_duration_max_seconds seconds
eventstore_queue_queueing_duration_max_seconds{name="Readers",range="16-20 seconds"} 0.06434454 1688157489545
eventstore_queue_queueing_duration_max_seconds{name="Others",range="16-20 seconds"} 0 1688157489545

Status

EventStoreDB tracks the current status of the Node role as well as progress of Index, and Scavenge processes.

Time series	Type	Description
`eventstore_statuses{name=<NAME>,status=<STATUS>}`	Gauge	Number of seconds since the 1970 epoch when NAME most recently had the status STATUS

For a given NAME, the current status can be determined by taking the max of all the time series with that name.

Index can have one of the following statuses:

Opening (loading/verifying the PTables)
Rebuilding (indexing previously written records on start up)
Initializing (initializing any other parts of the index e.g. StreamExistenceFilter on start up)
Merging
Scavenging
Idle

Scavenge can have one of the following statuses:

Accumulation
Calculation
Chunk Execution
Chunk Merging
Index Execution
Cleaning
Idle

Node can be one of the node roles.

Example configuration:

"Statuses": {
  "Index": true,
  "Node": false,
  "Scavenge": false
}

Example output:

# TYPE eventstore_checkpoints gauge
eventstore_statuses{name="Index",status="Idle"} 1688054162 1688054162477

Storage Writer

Time series	Type	Description
`eventstore_writer_flush_size_max{range=<RANGE>}`	RecentMax	Recent maximum flush size in bytes
`eventstore_writer_flush_duration_max_seconds{range=<RANGE>}`	RecentMax	Recent maximum flush duration in seconds

Example configuration:

"Writer": {
  "FlushSize": true,
  "FlushDuration": false
}

Example output:

# TYPE eventstore_writer_flush_size_max gauge
eventstore_writer_flush_size_max{range="16-20 seconds"} 410 1688056823193

System

Time series	Type	Description
`eventstore_sys_load_avg{period=<"1m"\|"5m"\|"15m">}`	Gauge	Average system load in last `1`, `5`, and `15` minutes. This metric is only available for Unix-like systems
`eventstore_sys_cpu`	Gauge	Current CPU usage in percentage. This metric is unavailable for Unix-like systems
`eventstore_sys_mem_bytes{kind=<"free"\|"total">}`	Gauge	Current free/total memory in bytes
`eventstore_sys_disk_bytes{disk=<MOUNT_POINT>,kind=<"used"\|"total">}`	Gauge	Current used/total bytes of disk mounted at MOUNT_POINT

Example configuration:

"System": {
  "Cpu": false,
  "LoadAverage1m": false,
  "LoadAverage5m": false,
  "LoadAverage15m": false,
  "FreeMem": false,
  "TotalMem": false,
  "DriveTotalBytes": false,
  "DriveUsedBytes": true
}

Example output:

# TYPE eventstore_sys_disk_bytes gauge
# UNIT eventstore_sys_disk_bytes bytes
eventstore_sys_disk_bytes{disk="/home",kind="used"} 38947205120 1688070655500

Metric types

Common types

Please refer to Prometheus documentation for explanation of common metric types (Gauge, Counter and Histogram).

RecentMax

A gauge whose value represents the maximum out of a set of recent measurements. Its purpose is to capture spikes that would otherwise have fallen in-between scrapes.

Note

The ExpectedScrapeIntervalSeconds setting within metricsconfig.json can be used to control the size of the window that the max is calculated over. It represents the expected interval between scrapes by a consumer such as Prometheus. It can only take specific values: 0, 1, 5, 10 or multiples of 15.

Setting the expected scape interval correctly ensures that spikes in the time series will be captured by at least one scrape and at most two.

Example output: Following metric is reported when ExpectedScrapeIntervalSeconds is set to 15 seconds

# TYPE eventstore_writer_flush_size_max gauge
eventstore_writer_flush_size_max{range="16-20 seconds"} 1854 1688070655500

In above example, maximum reported is 1854. It is not a maximum measurement in last 15s but rather maximum measurement in last 16 to last 20 seconds i.e. the maximum measurement could have been recorded in last 16s, last 17s, …, upto last 20s.