Metrics Reference for mongot

For the complete MongoDB documentation index, see www.mongodb.com/docs/llms.txt

Metrics Reference for mongot

The mongot process exposes Prometheus metrics that describe its runtime health and performance across core areas of operation. This reference page describes key metrics that are relevant to day-to-day monitoring and troubleshooting. For the complete metric set, scrape the mongot Prometheus metrics endpoint at http://<mongot-host>:9946/metrics.

View Metrics

To view the raw metrics that mongot exposes, send an HTTP GET request to the following mongot Prometheus metrics endpoint:

http://<mongot-host>:9946/metrics

In this endpoint:

• <mongot-host> is the hostname or IP address of the mongot process.

• 9946 is the default port for the metrics endpoint. To configure the metrics endpoint port, see the metrics.address setting in the mongot configuration file.

• /metrics is the path for the metrics endpoint.

The /metrics endpoint returns metrics in plain Prometheus text format. To monitor mongot metrics over time, configure your Prometheus instance to scrape this endpoint.

Important:

The /metrics endpoint requires no authentication by default. For production deployments, restrict access at the network layer.

Metric Naming Conventions

mongot metric names use a consistent naming pattern:

• All metric names start with the mongot_ prefix.

• Metric names generally follow the pattern mongot_<area>_<measurement>[_<unit>], where:

• <area> indicates the subsystem or component the metric belongs to, such as process, jvm, replication, or index.

• <measurement> indicates what is being measured, such as cpu_usage, heap_memory, or index_size.

• <unit> (optional) indicates the unit or counter semantics for the metric. This optional suffix indicates either the unit that the metric is measured in, such as seconds, bytes, or ms, or the type of counter the metric represents, such as total, events, or operations.

  Note:

  Some metric name suffixes don't reflect the actual reported unit for the metric. For example, mongot_index_stats_query_latency_seconds has the suffix _seconds, but mongot reports the metric in milliseconds, as indicated by the timeUnit=milliseconds label in the metric output. To confirm the unit for each metric, check the Unit value in the metric reference tables below.

In addition to the metric name, mongot metrics can include labels (also called dimensions). Labels distinguish multiple time series that share the same base metric name. For example, a metric might use labels to identify a state, status, index type, quantile, or a specific index.

For some metrics, you must interpret the metric as the combination of the metric name and its labels, not by the metric name alone. For example, mongot_replication_mongodb_indexManagerState uses the state label to expose one series for each replication state, such as STEADY_STATE or FAILED. Exactly one of those labeled series has the value 1 at a time. Per-index metrics similarly use labels such as generationId_logString and indexId_logString to distinguish one index from another.

For distribution metrics, the suffix of the metric name indicates the Prometheus series type:

• Histograms expose _bucket, _count, _sum, and _max.

• Summaries expose _count, _sum, and _max. Some summaries also include quantile labels such as {quantile=“0.5”}.

Common Metric Labels

The following table describes common labels that appear in mongot metrics.

Key Metric Groups

Process and JVM Metrics

Use process and JVM (Java Virtual Machine) metrics to confirm that mongot is running normally and to identify heap or garbage collection pressure.

Process

JVM Memory

JVM Garbage Collection

System Metrics

Use system metrics to monitor host-level CPU, disk, memory, paging, and network conditions that can affect mongot.

CPU

Disk

Memory

Page Faults

Network

Replication Metrics

Use replication metrics to determine whether mongot is healthy, syncing normally, and staying caught up with mongod.

Overall State

Session Refresher

Optime Updater

Per-index Metrics

mongot emits the following metrics per index and includes generationId_logString and indexId_logString labels to identify the specific index. Filter by those labels to inspect a specific index, or aggregate across labels to understand fleet-wide behavior.

Index Status and Size

Indexing Metrics

Query Metrics

Per-index Replication Breakdown

Lucene Refresh Latency

Command Metrics

mongot accepts the following set of named commands from mongod:

• buildinfo

• getMore

• hello

• isMaster

• ismaster

• killCursors

• manageSearchIndex

• ping

• planShardedSearch

• search

• vectorSearch

For each command, mongot exposes the following metrics, where <name> is a placeholder for the command name:

Tip: Monitor Search and Vector Search Latency Across Indexes

mongot_command_searchCommandTotalLatency_seconds and mongot_command_vectorSearchCommandTotalLatency_seconds are the primary metrics to monitor for [$search](https://www.mongodb.com/docs/search/query/aggregation-stages/search.md#mongodb-pipeline-pipe.-search and [$vectorSearch](https://www.mongodb.com/docs/vector-search/query/aggregation-stages/vector-search-stage.md#mongodb-pipeline-pipe.-vectorSearch latency aggregates. These expose latency aggregates for all search and vectorSearch commands across all indexes.

Indexing Scheduler and Dispatcher Metrics

Use indexing scheduler and dispatcher metrics to identify backlog, saturation, and slow work in replication and indexing pipelines.

Steady-state Change-stream Pipeline

Indexing Work Scheduler

Indexing work scheduler metrics monitor the queueing and execution of indexing batches.

Decoding Work Scheduler

Decoding work scheduler metrics monitor the queueing and execution of change-stream batch decoding.

Initial Sync, Lifecycle, and Config Metrics

Use these metrics to track index startup work, recovery, and catalog state.

Initial Sync

Note:

Some mongot metrics are phase-specific and populate only when the corresponding code path is active. For example, steady-state replication metrics, such as mongot_index_stats_indexing_replicationLagMs and the mongot_index_stats_replication_steadyState_* series, do not populate while an index is in initial sync. Conversely, initial-sync-specific metrics, such as mongot_initialsync_* and mongot_index_stats_replication_initialSync_*, are only relevant while initial sync is running or has run.

Lifecycle

Config State

Cursors and Index Factory

Use these metrics to monitor open cursor state and to detect indexes that mongot dropped or recovered because their on-disk segments were unreadable.

Lucene Merge

Use these metrics to monitor Lucene segment merge activity, including the number and size of merges in progress, merge input and output sizes, merge durations, and merges discarded by the disk-utilization merge policy.

MongoDB Client Connection Pool Metrics

mongot opens multiple named connection pools to mongod, and labels each pool with a clientName label that identifies the role of each pool. The following table lists possible clientName label values and their corresponding role:

The following table lists the available metrics for mongot connection pools:

Synonyms

Use these metrics to monitor synonym synchronization activity, including collection scans, scan and sync durations, queue depth, and exceptions encountered during synonym sync.

Executor Pools

Use these metrics to monitor the named executor pools that mongot uses to run background work. Each pool exposes the same set of sub-metrics, prefixed with the pool name, so you can track thread activity, pool sizing, queue depth, task throughput, and per-task execution time across all pools.

The following table lists the sub-metrics that every executor pool exposes, where <pool> is the pool name prefix. All executor-pool sub-metrics carry the label name=“executorMetrics”.

The following table lists the prefixes for all available named executor pools and their respective purposes:

Tip: Watch Saturation Across All Executor Pools

To monitor saturation across all executor pools, run the following PromQL query:

max by (pool) (
  label_replace(
    {__name__=~"mongot_.+_executor_queued_tasks"},
    "pool", "$1", "__name__", "mongot_(.+)_executor_queued_tasks"
  )
)

This query returns the queued-task count for each executor pool.

Prometheus Server Self-Metrics

The following metric is available for the embedded Prometheus server in mongot:

Configuration-Specific Metrics

The following metric families appear in the /metrics output only when you enable specific features.

PromQL Examples

Most latency metrics in this catalog are summaries, not histograms, so use their published quantile labels directly when they exist. A smaller number of metrics, such as mongot_index_stats_query_limitPerQuery and mongot_index_stats_query_numCandidatesPerQuery, are histograms and expose _bucket series.

## Replication state
max by (state) (mongot_replication_mongodb_indexManagerState)

## Number of search indexes at STEADY_STATE replication
mongot_replication_mongodb_indexManagerState{state="STEADY_STATE"}

## Maximum replication lag across all indexes, converted to seconds
max(mongot_index_stats_indexing_replicationLagMs) / 1000

## Index count by status
count by (status) (mongot_index_stats_indexStatusCode == 1)

## Search query p99 latency across all indexes
max(mongot_index_stats_query_searchResultBatchLatencies_seconds{quantile="0.99"})

## Worst recent GC pause
max(mongot_jvm_gc_pause_seconds_max)

## Average GC pause over 5 minutes
rate(mongot_jvm_gc_pause_seconds_sum[5m]) / rate(mongot_jvm_gc_pause_seconds_count[5m])

## Free disk percentage on dataPath
mongot_system_disk_space_data_path_free_bytes
  / mongot_system_disk_space_data_path_total_bytes

## Major page fault rate
rate(mongot_system_process_majorPageFaults_operations[5m])

## Steady-state and initial sync exceptions over 15 minutes
sum(rate(mongot_index_stats_indexing_steadyStateExceptions_total[15m]))
sum(rate(mongot_index_stats_indexing_initialSyncExceptions_total[15m]))