Review Available Metrics

Displays the following information:

• ASSERT_REGULAR displays the average rate of regular asserts raised per second over the selected sample period.

• ASSERT_WARNING displays the average rate of warnings per second over the selected sample period.

• ASSERT_MSG displays the average rate of message asserts per second over the selected sample period. These internal server errors have a well-defined text string. Atlas logs stack traces for these.

• ASSERT_USER displays the average rate of user asserts per second over the selected sample period. This metric includes asserts that a user generates, such as out of disk space or duplicate key errors.

Monitor asserts to track how many errors occur while trying to read or write data. Check the server logs to identify the source of any errors.

Displays the average object size across all collections in the database.

Monitor object size to track the size of your objects and better understand your database space.

Displays the following information:

• readInto (replica set) or cache read into (sharded cluster) displays the rate in bytes per second of data read from disk into memory to service queries.

• writtenFrom (replica set) or cache written from (sharded cluster) displays the rate in bytes per second of data written from memory into disk to service writes.

Monitor the MongoDB cache, which stores frequently accessed data in memory to service queries faster.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays cache fill ratio and dirty fill ratio metrics.

Cache Fill Ratio measures how well a cache can serve requests. It is calculated by dividing the number of bytes currently in the cache by the maximum number of bytes configured, represented as a percentage. A high cache fill ratio indicates that most data requests are being served from memory, leading to faster query performance and reduced disk I/O.

Dirty Fill Ratio represents the proportion of dirty bytes, which are pages modified in memory but not yet written to disk, relative to the total cache. A high dirty fill ratio suggests that a significant amount of data is waiting to be flushed to disk, which can impact performance. Use this metric when monitoring write-heavy workloads to ensure data durability.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the following information:

• dirty (replica set) or cache dirty (sharded cluster) displays the total dirty bytes cached in memory for serving reads and writes.

• used (replica set) or cache used (sharded cluster) displays the total bytes cached in memory for serving reads and writes.

These metrics include both indexes and data from the working set.

Sustained high cache usage indicates the RAM is too small for your workloads. Optimize your queries to avoid frequent disk reads. If write operations make cache usage high, throttle them.

Displays the amount of cache pressure currently experienced by the node, expressed as a percentage and capped at 100%. This metric is available for MongoDB 9.0 or later.

Displays the following information:

• total databases displays the total number of non-system databases.

• total collections displays the total number of collections across all non-system databases.

• total views displays the total number of views across all non-system databases.

• total indexes displays the total number of indexes across all non-system databases.

• total namespaces displays the total number of collections and indexes across all non-system databases.

Monitor catalog counts to prevent an excessive number of databases, collections, views, indexes, or namespaces from causing startup failures when you scale a cluster tier.

Displays the number of collections in the database.

Monitor collections to determine restart times, continuous backup performance, and stability.

Displays the following information about connection rate limiting metrics for Intelligent Workload Management:

• Offered Connections displays the average rate of connection requests per second offered to the ingress session establishment queue.

• Successful Connections displays the average rate of connection requests per second that successfully pass through the rate limiter.

• Rejected Connections displays the average rate of connection requests per second that are rejected because the connection rate limit was exceeded.

• Interrupted In Queue displays the average rate of connection requests per second that are interrupted while waiting in the ingress session establishment queue, typically because of client disconnects or timeouts.

• Average Time Queued displays the average time in microseconds that connection requests spend waiting in the ingress session establishment queue before admission.

Monitor connection rate limiting metrics to understand how MongoDB manages connection establishment during high-load conditions. A high rejection rate indicates that the cluster is protecting itself from overload and may benefit from throttling clients, reducing connection pool sizes, or scaling up.

Connections (replica set) or connection (sharded cluster)

Displays the total number of active connections to the cluster.

Monitor connections to determine whether the current connection limits are sufficient. If necessary, scale the cluster tier.

Displays the following information:

• totalOpen displays the number of cursors that the server maintains for clients.

• totalTimedOut displays the average rate of cursors that have timed out per second over the selected sample period.

Monitor cursors to close unnecessary cursors and reduce the timeout configuration in the application.

Displays the following information:

• dataSize (replica set), or db data size (sharded cluster) displays the logical size of uncompressed data.

• db data size without system displays the sum total size in bytes of the document data (including the padding factor) across non-system databases.

• db index size displays the sum of the disk space allocated to all indexes in the database, including free index space.

Atlas retrieves database metrics every 20 minutes by default but adjusts frequency when necessary to reduce the impact on database performance.

Monitor storage space to determine whether to use disk auto-scaling or manually increase the disk size. You can also monitor this metric to verify backup billing.

Displays input operations per second.

Monitor whether disk IOPS approaches the maximum provisioned IOPS. Determine whether the cluster can handle future workloads.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the following information:

• Read displays the average amount of time to read from disk.

• Write displays the average amount of time to write to disk.

Monitor disk latency to track the efficiency of reading from and writing to disk.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the average length of the queue of requests issued to the disk partition that MongoDB uses.

Monitor disk queue depth to identify potential issues and bottlenecks.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the total amount of free space remaining on disk.

Monitor free disk space to determine whether to use disk auto-scaling or manually increase the disk size.

IMPORTANT: The UI displays GB, but all disk usage metrics are in gibibytes (GiB).

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the total amount of free space remaining on disk as a percentage of the total disk space.

Monitor the percentage of free disk space to determine whether to use disk auto-scaling or manually increase the disk size.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the total space on disk used.

Monitor the used disk space to determine whether to use disk auto-scaling or manually increase the disk size.

If Atlas identifies that your cluster's primary node exceeds write-blocking thresholds, Atlas blocks writes to the cluster until disk usage falls below the unblocking thresholds.

IMPORTANT: The UI displays GB, but all disk usage metrics are in gibibytes (GiB).

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the disk read and write throughput metrics.

Disk Read Throughput reflects the rate at which data is read from disk in Megabytes per second, indicating how efficiently the database retrieves data that is not cached in memory.

Disk Write Throughput measures the speed at which data is written to disk in Megabytes per second, reflecting the database's ability to handle write operations and persist data to storage efficiently.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the following information:

• Returned displays the documents per second returned.

• Inserted displays the documents per second inserted.

• Updated displays the documents per second updated.

• Deleted displays the documents per second deleted.

Monitor document metrics to measure the work MongoDB completes.

Displays the average time in seconds for the following metrics:

• Average read operational latency
• Average write operational latency
• Average command operational latency

Monitor execution time for an increase in read operations to optimize queries and indexes.

Displays the total size of all indexes in the database. This metric includes the overhead incurred by indexes on top of the actual document data on which the indexes are based.

Monitor the index size to manage your indexes. To learn more, see Indexing Strategies.

Displays the total number of indexes in the database.

Monitor indexes to manage them. To learn more, see Indexing Strategies.

Displays the following maximum disk IOPS values over the time period specified by the metric granularity:

• max read iops maximum disk read input operations per second.

• max write iops maximum disk write input operations per second.

Monitor whether disk IOPS approaches the maximum provisioned IOPS. Determine whether the cluster can handle future workloads.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the maximum disk queue depth values over the time period specified by the metric granularity. Disk queue depth is the average length of the queue of requests issued to the disk partition that MongoDB uses.

Monitor disk queue depth to identify potential issues and bottlenecks.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the maximum CPU usage values of all processes on the node, scaled to a range of 0-100% by dividing by the number of CPU cores.

Monitor CPU usage to determine whether to scale to a higher cluster tier.

If you are unable to see the usage that triggered the alert, zoom in on the Normalized System CPU chart by clicking and dragging your mouse over the period of interest. With a higher-resolution view you may be able to identify acute spikes in CPU usage that weren't visible in the overview.

Displays the following maximum process CPU values over the time period specified by the metric granularity:

• max user displays the maximum percentage of time that the CPU spent servicing the MongoDB process.

• max kernel displays the maximum percentage of time the CPU spent servicing operating system calls for the MongoDB process.

Monitor CPU usage to determine whether to scale to a higher cluster tier.

If you are unable to see the usage that triggered the alert, zoom in on the Normalized System CPU chart by clicking and dragging your mouse over the period of interest. With a higher-resolution view you may be able to identify acute spikes in CPU usage that weren't visible in the overview.

Displays the maximum CPU usage values of all processes on the node.

Monitor CPU usage to determine whether to scale to a higher cluster tier.

If you are unable to see the usage that triggered the alert, zoom in on the Normalized System CPU chart by clicking and dragging your mouse over the period of interest. With a higher-resolution view you may be able to identify acute spikes in CPU usage that weren't visible in the overview.

Displays the maximum system memory values in bytes.

Monitor memory to determine whether to scale to a higher cluster tier. This metric is based on free memory value, where MEM USED value includes reclaimable system memory.

Displays the amount of memory in use at a particular point in time, measured in megabytes.

• memory_resident or memory resident displays the amount of physical memory that the MongoDB process running on a node consumes. This metric excludes the memory consumption by processes other than MongoDB running on a node. This metric doesn't represent the total memory that the node consumes.

• memory_virtual or memory virtual displays the amount of virtual memory for the MongoDB process running on a node.

Monitor memory to determine whether to scale to a higher cluster tier. This metric represents the average value over the time period specified by the metric granularity.

Displays the following information:

• bytesIn displays the average rate of physical bytes (after any wire compression) sent to this database server per second over the selected sample period.

• bytesOut displays the average rate of physical bytes (after any wire compression) sent from this database server per second over the selected sample period.

• numRequests displays the average rate of requests sent to this database server per second over the selected sample period.

  Monitor network metrics to track network performance.

Displays the following information:

• user displays the percentage of time that the CPU spent servicing the MongoDB process, scaled to a range of 0-100% by dividing by the number of CPU cores.

• kernel displays the percentage of time the CPU spent servicing operating system calls for the MongoDB process, scaled to a range of 0-100% by dividing by the number of CPU cores.

Monitor CPU usage to determine whether to scale to a higher cluster tier.

If you are unable to see the usage that triggered the alert, zoom in on the Normalized System CPU chart by clicking and dragging your mouse over the period of interest. With a higher-resolution view you may be able to identify acute spikes in CPU usage that weren't visible in the overview.

Displays the CPU usage of all processes on the node, scaled to a range of 0-100% by dividing by the number of CPU cores.

Monitor CPU usage to determine whether to scale to a higher cluster tier.

If you are unable to see the usage that triggered the alert, zoom in on the Normalized System CPU chart by clicking and dragging your mouse over the period of interest. With a higher-resolution view you may be able to identify acute spikes in CPU usage that weren't visible in the overview.

Displays the number of objects in the database.

Monitor this metric to better understand your database space.

Displays the number of the following operations per second run on a MongoDB process since the process last started:

• command (replica set) or cmd (sharded cluster)

• query
• insert
• delete
• update
• getmore
• ttldeleted

ttldeleted from TTL indexes only occur against a node in a PRIMARY state and are replicated as standard deletes to the secondary nodes. Because of this, TTL deletes aren't applicable on secondaries. Atlas maintains the history of the previous primary member's TTL deletes and the TTL deletes remain visible on secondaries that had previously served as the primary member.

Monitor MongoDB operations to validate performance issues related to high workloads. Confirm the type of operations responsible for the load.

Displays the following information:

• command displays the average rate of replicated commands applied per second over the selected sample period.

• insert displays the average rate of replicated inserts applied per second over the selected sample period.

• delete displays the average rate of replicated deletes applied per second over the selected sample period.

• update displays the average rate of replicated updates applied per second over the selected sample period.

Monitor MongoDB operations to validate performance issues related to high workloads. Confirm the type of operations responsible for the load.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the average time in milliseconds to execute the following operations:

• avg ms/read (replica set) or execution time reads (sharded cluster)

• avg ms/write (replica set) or execution time writes (sharded cluster)

• avg ms/command (replica set) or execution time commands (sharded cluster)

Monitor execution time for an increase in read operations to optimize queries and indexes. Determine whether you need to scale your cluster tier.

Displays the following information about operation rate limiting metrics for Intelligent Workload Management:

• Offered Operations displays the average rate of operations per second offered to the ingress request rate limiter.

• Successful Operations displays the average rate of operations per second that successfully pass through the rate limiter and are admitted for execution.

• Rejected Operations displays the average rate of operations per second that are rejected by the rate limiter because the system is overloaded.

• Operation Queue Time (Avg) displays the average time in microseconds that operations spend waiting in the ingress request rate limiter queue before admission.

• Queued Operations displays the rate of operations added to the ingress request rate limiter queue.

• Current Queue Depth displays the current number of operations waiting in the ingress request rate limiter queue.

Monitor operation rate limiting metrics to understand how MongoDB manages incoming operations during high-load conditions. A high rejection rate indicates that the cluster is protecting itself from overload and may benefit from optimizing queries, reducing background work, or scaling up.

Displays the average time in milliseconds that operations wait in the ingress admission queue before being processed. This metric is available for MongoDB 9.0 or later.

Displays the following information:

Monitor operation throttling to understand how your application workload impacts your cluster.

Displays the average rate of the uncompressed oplog data in gigabytes that the primary generates per hour.

Monitor oplog data to determine whether you have to increase the oplog size.

Number of orphaned documents in the shard.

Size in bytes of orphaned documents in the shard when uncompressed.

Displays the average rate of page faults on this process per second over the selected sample period. In non-Windows environments this applies to hard page faults only.

Monitor page faults to determine whether to increase your memory.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the following information:

• user displays the percentage of time that the CPU spent servicing the MongoDB process.

• kernel displays the percentage of time the CPU spent servicing operating system calls for the MongoDB process.

Monitor CPU usage to determine whether to scale to a higher cluster tier.

If you are unable to see the usage that triggered the alert, zoom in on the Normalized System CPU chart by clicking and dragging your mouse over the period of interest. With a higher-resolution view you may be able to identify acute spikes in CPU usage that weren't visible in the overview.

Displays the following information:

• Index Items Scanned displays the number of index items scanned per second.

• Documents Scanned displays the number of documents scanned per second.

Monitor the query executor to determine whether you have any inefficient queries.

Displays the efficiency of read operations run on MongoDB:

• Scanned Objects to Returned (replica set) or scanned objects / returned (sharded cluster) displays the number of documents scanned to return one document.

• Scanned Keys to Returned (replica set) or scanned keys / returned (sharded cluster) displays the number of index keys scanned to return one document.

Monitor query targeting to determine read efficiency and optimize queries and indexes.

The change streams cursors that the MongoDB Search process (mongot) uses to keep MongoDB Search indexes updated can contribute to the query targeting ratio and trigger query targeting alerts if the ratio is high.

Use this series to view the amount of data spilled to disk by queries. This metric is available for MongoDB 9.0 or later.

• File Spilled Size displays the amount of storage in bytes used by query operations that spilled to disk.

Displays the following information:

• GLOBAL_LOCK_CURRENT_QUEUE_TOTAL displays the number of operations queued waiting for any lock.

• GLOBAL_LOCK_CURRENT_QUEUE_READERS displays the number of operations queued waiting for a read lock.

• GLOBAL_LOCK_CURRENT_QUEUE_WRITERS displays the number of operations queued waiting for a write lock.

Monitor lock queues to optimize queries.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the following information:

• Total Read Processing Units (RPUs)

• Total Write Processing Units (WPUs)

Monitor read and write units to help optimize queries and indexes.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the difference between the primary's replication oplog window and the secondary's replication lag.

Monitor replication headroom to determine whether the secondary might fall off the oplog.

Atlas collects this data at 85-second intervals regardless of your project's default granularity level.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the approximate number of seconds the secondary is behind the primary in write application.

For Atlas Core, monitor replication lag to determine whether the secondary might fall off the oplog. For Atlas Infinite, this metric shows the lag for writes to become visible on the node.

Atlas collects this data at 85-second intervals regardless of your project's default granularity level.

For Atlas Core, displays the estimated average number of hours of database operations available in the primary's replication oplog, based on oplog churn. If replication lag on a secondary node exceeds the replication oplog window, and replication headroom reaches zero, a full resync is required for that node to become healthy again.

For Atlas Infinite, this metric measures the number of hours of recent changes the database has stored for features such as change streams. To learn more, see MongoDB Atlas Infinite: Architecture.

Monitor the replication oplog window, together with replication headroom, to determine whether the secondary may soon require a full resync. The replication oplog window often helps to determine in advance the resilience of secondaries to planned and unplanned outages.

Displays the number of operations per second returning results that required a sort in-memory.

Monitor this metric to identify whether your queries need indexes.

Displays the amount of storage space in bytes that your stored data uses on each shard.

Monitor this metric to verify whether you have balanced shards.

Displays the number of documents on each shard.

Monitor this metric to verify whether you have balanced shards.

Displays spill to disk during sort, the total number of writes to disk caused by $sort stages.

A high spill to disk during sort number might be due to resource-intensive queries that are performing a collection scan and are using a $sort stage. These resource-intensive queries result in temporary file creation that exceed available memory and use disk space for sorting. Identify offending queries in the Query Profiler by looking for collection scans with a $sort stage around the same time the database experiences an increase in disk usage.

Displays the following information about swap space on the host:

• used displays the total amount of swap space currently in use.

• free displays the total amount of swap space that is still available.

• max used displays the maximum total amount of swap space in use over the selected time period.

• max free displays the maximum amount of available swap space over the selected time period.

Sustained high swap usage can indicate memory pressure on the host.

Monitor swap usage to identify when the operating system starts moving memory pages to disk.

Displays the CPU usage of all processes on the node.

Monitor CPU usage to determine whether to scale to a higher cluster tier.

If you are unable to see the usage that triggered the alert, zoom in on the Normalized System CPU chart by clicking and dragging your mouse over the period of interest. With a higher-resolution view you may be able to identify acute spikes in CPU usage that weren't visible in the overview.

Displays the following information:

• Memory Used (bytes): Displays the number of bytes of physical memory in use on the host. Atlas calculates Memory Used as memoryTotal - (memoryFree + memoryBuffers + memoryCached), where memoryFree, memoryBuffers, and memoryCached are amounts of available memory that Atlas can reclaim for other purposes.

• Memory Used (percent): The percentage of total physical memory in use on the host. Derived from (MemTotal - MemAvailable) / MemTotal. For modern Linux kernels, MemAvailable is stable and a better heuristic than the MemFree + Buffers + Cached calculation made in the "Memory Used (Bytes)" series. Derived from MemTotal - (MemFree + Buffers + Cached).

• Memory Available: An estimate of the number of bytes of system memory available for running new applications, without swapping.

• Max Memory Used (bytes): The maximum memory used in bytes over the sampling period.

• Max Memory Used (percent): The maximum memory used as a percentage of total memory over the sampling period.

• Max Memory Available: The maximum memory available in bytes over the sampling period.

Monitor memory to determine whether to scale to a higher cluster tier. This metric represents the average value over the time period specified by the metric granularity.

Displays the following information:

• network bytes in displays the average rate of physical bytes that the network interface receives per second.

• network bytes out displays the average rate of physical bytes that the network interface transmits per second.

• max network bytes in displays the maximum network bytes in values over the time period that the metric granularity specifies.

• max network bytes out displays the maximum network bytes out values over the time period that the metric granularity specifies.

Monitor network metrics to track network performance.

Use these series to view multi-document transaction metrics for your MongoDB process(es):

• Current Active displays the number of transactions currently active i.e., executing an operation.

• Current Inactive displays the number of open transactions not currently executing an operation.

• Current Open displays the total number of transactions currently open.

• Total Aborted displays the average rate of transactions aborted per second.

• Total Committed displays the average rate of transactions committed per second.

• Total Started displays the average rate of transactions started per second.

Displays the following information:

• Tickets Available Read displays the number of read tickets available to the WiredTiger storage engine. Read tickets represent the number of concurrent read operations allowed into the storage engine. When this value reaches zero, new read requests might queue until a read ticket becomes available.

• Tickets Available Write displays the number of write tickets available to the WiredTiger storage engine. Write tickets represent the number of concurrent write operations allowed into the storage engine. When this value reaches zero, new write requests might queue until a write ticket becomes available.

Monitor the tickets available to see when read and write requests queue.

For clusters running on MongoDB version 7.0 and later, don't use the number of tickets as a metric for overload alerts. Starting in MongoDB version 7.0, Atlas dynamically adjusts the number of tickets. Instead, use the number of queued readers and writers as an overload metric.

Atlas Core only: This metric measures node-level performance and is not available for Atlas Infinite. Atlas Infinite uses a decoupled storage architecture with different metrics. To learn more, see Atlas Infinite Metrics.

Displays the number of views in the database.

Monitor views to help optimize your database.

Storage IOPS

Use these series to view I/O operations per second against Atlas Infinite's storage layer:

• Read IOPS displays the average rate of read operations per second.

• Write IOPS displays the average rate of write operations per second.

Storage Latency

Use these series to view the read and write latency for operations against Atlas Infinite's storage layer:

• Read Latency displays the average latency of reads in milliseconds from Atlas Infinite's storage layer.

• Write Latency displays the average latency of writes in milliseconds to Atlas Infinite's storage layer.

Storage Throughput

Use these series to view bytes read from and written to Atlas Infinite's storage layer:

• Bytes Read displays the average rate, in bytes-per-second, of data read from Atlas Infinite's storage layer.

• Bytes Written displays the average rate, in bytes-per-second, of data written to Atlas Infinite's storage layer.

Storage Throughput Utilization

Displays the percentage of consumed throughput to Atlas Infinite's storage layer.

Storage Space

Use these series to view storage space usage for Atlas Infinite:

• Storage Space Free displays the amount of free storage space in bytes.

• Storage Space Percent Free displays the percentage of storage space that is free.

• Storage Space Used displays the amount of storage space in bytes that is currently in use.

• Storage Space Percent Used displays the percentage of storage space that is in use.

Active Collections and Indexes

Displays the number of active collections and indexes.

Avg Majority Write Concern Wait

The average time write operations in milliseconds wait to satisfy their majority write concern. Elevated values indicate replication acknowledgement is slow.