Docs Menu
Docs Home
/
MongoDB Manual
/

Log Messages

On this page

  • Overview
  • Structured Logging
  • Log Message Field Types
  • Verbosity Levels
  • Log Redaction
  • Parsing Structured Log Messages
  • Log Message Examples
  • Linux Syslog Limitations
  • Download Your Logs

As part of normal operation, MongoDB maintains a running log of events, including entries such as incoming connections, commands run, and issues encountered. Generally, log messages are useful for diagnosing issues, monitoring your deployment, and tuning performance.

To get your log messages, you can use any of the following methods:

mongod / mongos instances output all log messages in structured JSON format. Log entries are written as a series of key-value pairs, where each key indicates a log message field type, such as "severity", and each corresponding value records the associated logging information for that field type, such as "informational". Previously, log entries were output as plaintext.

Example

The following is an example log message in JSON format as it would appear in the MongoDB log file:

{"t":{"$date":"2020-05-01T15:16:17.180+00:00"},"s":"I", "c":"NETWORK", "id":12345, "ctx":"listener", "svc": "R", "msg":"Listening on", "attr":{"address":"127.0.0.1"}}

JSON log entries can be pretty-printed for readability. Here is the same log entry pretty-printed:

{
"t": {
"$date": "2020-05-01T15:16:17.180+00:00"
},
"s": "I",
"c": "NETWORK",
"id": 12345,
"ctx": "listener",
"svc": "R",
"msg": "Listening on",
"attr": {
"address": "127.0.0.1"
}
}

In this log entry, for example, the key s, representing severity, has a corresponding value of I, representing "Informational", and the key c, representing component, has a corresponding value of NETWORK, indicating that the "network" component was responsible for this particular message. The various field types are presented in detail in the Log Message Field Types section.

Structured logging with key-value pairs allows for efficient parsing by automated tools or log ingestion services, and makes programmatic search and analysis of log messages easier to perform. Examples of analyzing structured log messages can be found in the Parsing Structured Log Messages section.

Note

The mongod quits if it's unable to write to the log file. To ensure that mongod can write to the log file, verify that the log volume has space on the disk and the logs are rotated.

All log output is in JSON format including output sent to:

Output from the getLog command is also in JSON format.

Each log entry is output as a self-contained JSON object which follows the Relaxed Extended JSON v2.0 specification, and has the following layout and field order:

{
"t": <Datetime>, // timestamp
"s": <String>, // severity
"c": <String>, // component
"id": <Integer>, // unique identifier
"ctx": <String>, // context
"svc": <String>, // service
"msg": <String>, // message body
"attr": <Object>, // additional attributes (optional)
"tags": <Array of strings>, // tags (optional)
"truncated": <Object>, // truncation info (if truncated)
"size": <Object> // original size of entry (if truncated)
}

Field descriptions:

Field Name
Type
Description
t
Datetime
Timestamp of the log message in ISO-8601 format. For an example, see Timestamp.
s
String
Short severity code of the log message. For an example, see Severity.
c
String
Full component string for the log message. For an example, see Components.
id
Integer
Unique identifier for the log statement. For an example, see Filtering by Known Log ID.
ctx
String
Name of the thread that caused the log statement.
svc
String
Name of the service in whose context the log statement was made. Will be S for "shard", R "router", or - for "unknown" or "none".
msg
String
Log output message passed from the server or driver. If necessary, the message is escaped according to the JSON specification.
attr
Object
One or more key-value pairs for additional log attributes. If a
log message does not include any additional attributes, the
attr object is omitted.
Attribute values may be referenced by their key name in the
msg message body, depending on the message. If necessary, the
attributes are escaped
according to the JSON specification.
tags
Array of strings
Strings representing any tags applicable to the log statement. For example, ["startupWarnings"].
truncated
Object
Information about the log message truncation, if applicable. Only included if the log entry contains at least one truncated attr attribute.
size
Object
Original size of a log entry if it has been truncated. Only included if the log entry contains at least one truncated attr attribute.

The message and attributes fields will escape control characters as necessary according to the Relaxed Extended JSON v2.0 specification:

Character Represented
Escape Sequence
Quotation Mark (")
\"
Backslash (\)
\\
Backspace (0x08)
\b
Formfeed (0x0C)
\f
Newline (0x0A)
\n
Carriage return (0x0D)
\r
Horizontal tab (0x09)
\t

Control characters not listed above are escaped with \uXXXX where "XXXX" is the unicode codepoint in hexadecimal. Bytes with invalid UTF-8 encoding are replaced with the unicode replacement character represented by \ufffd.

An example of message escaping is provided in the examples section.

Changed in version 7.3.

Any attributes that exceed the maximum size defined with maxLogSizeKB (default: 10 KB) are truncated. Truncated attributes omit log data beyond the configured limit, but retain the JSON formatting of the entry to ensure that the entry remains parsable.

For example, the following JSON object represents a command attribute that contains 5000 elements in the $in field without truncation.

Note

The example log entries are reformatted for readability.

{
"command": {
"find": "mycoll",
"filter": {
"value1": {
"$in": [0, 1, 2, 3, ... 4999]
},
"value2": "foo"
},
"sort": { "value1": 1 },
"lsid":{"id":{"$uuid":"80a99e49-a850-467b-a26d-aeb2d8b9f42b"}},
"$db": "testdb"
}
}

In this example, the $in array is truncated at the 376th element because the size of the command attribute would exceed maxLogSizeKB if it included the subsequent elements. The remainder of the command attribute is omitted. The truncated log entry resembles the following output:

{
"t": { "$date": "2021-03-17T20:30:07.212+01:00" },
"s": "I",
"c": "COMMAND",
"id": 51803,
"ctx": "conn9",
"msg": "Slow query",
"attr": {
"command": {
"find": "mycoll",
"filter": {
"value1": {
"$in": [ 0, 1, ..., 376 ] // Values in array omitted for brevity
}
}
},
... // Other attr fields omitted for brevity
},
"truncated": {
"command": {
"truncated": {
"filter": {
"truncated": {
"value1": {
"truncated": {
"$in": {
"truncated": {
"377": {
"type": "double",
"size": 8
}
},
"omitted": 4623
}
}
}
},
"omitted": 1
}
},
"omitted": 3
}
},
"size": {
"command": 21692
}
}

Log entries containing one or more truncated attributes include nested truncated objects, which provide the following information for each truncated attribute in the log entry:

  • The attribute that was truncated

  • The specific sub-object of that attribute that triggered truncation, if applicable

  • The data type of the truncated field

  • The size, in bytes, of the element that triggers truncation

  • The number of elements that were omitted under each sub-object due to truncation

Log entries with truncated attributes may also include an additional size field at the end of the entry which indicates the original size of the attribute before truncation, in this case 21692 or about 22KB. This final size field is only shown if it is different from the size field in the truncated object, i.e. if the total object size of the attribute is different from the size of the truncated sub-object, as is the case in the example above.

When output to the file or the syslog log destinations, padding is added after the severity, context, and id fields to increase readability when viewed with a fixed-width font.

The following MongoDB log file excerpt demonstrates this padding:

{"t":{"$date":"2020-05-18T20:18:12.724+00:00"},"s":"I", "c":"CONTROL", "id":23285, "ctx":"main", "svc": "R", "msg":"Automatically disabling TLS 1.0, to force-enable TLS 1.0 specify --sslDisabledProtocols 'none'"}
{"t":{"$date":"2020-05-18T20:18:12.734+00:00"},"s":"W", "c":"ASIO", "id":22601, "ctx":"main", "svc": "R", "msg":"No TransportLayer configured during NetworkInterface startup"}
{"t":{"$date":"2020-05-18T20:18:12.734+00:00"},"s":"I", "c":"NETWORK", "id":4648601, "ctx":"main", "svc": "R", "msg":"Implicit TCP FastOpen unavailable. If TCP FastOpen is required, set tcpFastOpenServer, tcpFastOpenClient, and tcpFastOpenQueueSize."}
{"t":{"$date":"2020-05-18T20:18:12.814+00:00"},"s":"I", "c":"STORAGE", "id":4615611, "ctx":"initandlisten", "svc": "R", "msg":"MongoDB starting", "attr":{"pid":10111,"port":27001,"dbPath":"/var/lib/mongo","architecture":"64-bit","host":"centos8"}}
{"t":{"$date":"2020-05-18T20:18:12.814+00:00"},"s":"I", "c":"CONTROL", "id":23403, "ctx":"initandlisten", "svc": "R", "msg":"Build Info", "attr":{"buildInfo":{"version":"4.4.0","gitVersion":"328c35e4b883540675fb4b626c53a08f74e43cf0","openSSLVersion":"OpenSSL 1.1.1c FIPS 28 May 2019","modules":[],"allocator":"tcmalloc","environment":{"distmod":"rhel80","distarch":"x86_64","target_arch":"x86_64"}}}}
{"t":{"$date":"2020-05-18T20:18:12.814+00:00"},"s":"I", "c":"CONTROL", "id":51765, "ctx":"initandlisten", "svc": "R", "msg":"Operating System", "attr":{"os":{"name":"CentOS Linux release 8.0.1905 (Core) ","version":"Kernel 4.18.0-80.11.2.el8_0.x86_64"}}}

When working with MongoDB structured logging, the third-party jq command-line utility is a useful tool that allows for easy pretty-printing of log entries, and powerful key-based matching and filtering.

jq is an open-source JSON parser, and is available for Linux, Windows, and macOS.

You can use jq to pretty-print log entries as follows:

  • Pretty-print the entire log file:

    cat mongod.log | jq
  • Pretty-print the most recent log entry:

    cat mongod.log | tail -1 | jq

More examples of working with MongoDB structured logs are available in the Parsing Structured Log Messages section.

MongoDB log messages can be output to file, syslog, or stdout (standard output).

To configure the log output destination, use one of the following settings, either in the configuration file or on the command-line:

Configuration file:
Command-line:

Not specifying either file or syslog sends all logging output to stdout.

For the full list of logging settings and options see:

Configuration file:
Command-line:

Note

Error messages sent to stderr (standard error), such as fatal errors during startup when not using the file or syslog log destinations, or messages having to do with misconfigured logging settings, are not affected by the log output destination setting, and are printed to stderr in plaintext format.

The timestamp field type indicates the precise date and time at which the logged event occurred.

{
"t": {
"$date": "2020-05-01T15:16:17.180+00:00"
},
"s": "I",
"c": "NETWORK",
"id": 12345,
"ctx": "listener",
"svc": "R",
"msg": "Listening on",
"attr": {
"address": "127.0.0.1"
}
}

When logging to file or to syslog [1], the default format for the timestamp is iso8601-local. To modify the timestamp format, use the --timeStampFormat runtime option or the systemLog.timeStampFormat setting.

See Filtering by Date Range for log parsing examples that filter on the timestamp field.

Note

The ctime timestamp format is no longer supported.

[1] If logging to syslog, the syslog daemon generates timestamps when it logs a message, not when MongoDB issues the message. This can lead to misleading timestamps for log entries, especially when the system is under heavy load.

The severity field type indicates the severity level associated with the logged event.

{
"t": {
"$date": "2020-05-01T15:16:17.180+00:00"
},
"s": "I",
"c": "NETWORK",
"id": 12345,
"ctx": "listener",
"svc": "R",
"msg": "Listening on",
"attr": {
"address": "127.0.0.1"
}
}

Severity levels range from "Fatal" (most severe) to "Debug" (least severe):

Level
Description
F
Fatal
E
Error
W
Warning
I
Informational, for verbosity level 0
D1 - D5

Debug, for verbosity levels > 0

MongoDB indicates the specific debug verbosity level. For example, if verbosity level is 2, MongoDB indicates D2.

In previous versions, MongoDB log messages specified D for all debug verbosity levels.

You can specify the verbosity level of various components to determine the amount of Informational and Debug messages MongoDB outputs. Severity categories above these levels are always shown. [2] To set verbosity levels, see Configure Log Verbosity Levels.

The component field type indicates the category a logged event is a member of, such as NETWORK or COMMAND.

{
"t": {
"$date": "2020-05-01T15:16:17.180+00:00"
},
"s": "I",
"c": "NETWORK",
"id": 12345,
"ctx": "listener",
"svc": "R",
"msg": "Listening on",
"attr": {
"address": "127.0.0.1"
}
}

Each component is individually configurable via its own verbosity filter. The available components are as follows:

ACCESS

Messages related to access control, such as authentication. To specify the log level for ACCESS components, use the systemLog.component.accessControl.verbosity setting.

COMMAND

Messages related to database commands, such as count. To specify the log level for COMMAND components, use the systemLog.component.command.verbosity setting.

CONTROL

Messages related to control activities, such as initialization. To specify the log level for CONTROL components, use the systemLog.component.control.verbosity setting.

ELECTION

Messages related specifically to replica set elections. To specify the log level for ELECTION components, set the systemLog.component.replication.election.verbosity parameter.

REPL is the parent component of ELECTION. If systemLog.component.replication.election.verbosity is unset, MongoDB uses the REPL verbosity level for ELECTION components.

FTDC

Messages related to the diagnostic data collection mechanism, such as server statistics and status messages. To specify the log level for FTDC components, use the systemLog.component.ftdc.verbosity setting.

GEO

Messages related to the parsing of geospatial shapes, such as verifying the GeoJSON shapes. To specify the log level for GEO components, set the systemLog.component.geo.verbosity parameter.

INDEX

Messages related to indexing operations, such as creating indexes. To specify the log level for INDEX components, set the systemLog.component.index.verbosity parameter.

INITSYNC

Messages related to initial sync operation. To specify the log level for INITSYNC components, set the systemLog.component.replication.initialSync.verbosity parameter.

REPL is the parent component of INITSYNC. If systemLog.component.replication.initialSync.verbosity is unset, MongoDB uses the REPL verbosity level for INITSYNC components.

JOURNAL

Messages related specifically to storage journaling activities. To specify the log level for JOURNAL components, use the systemLog.component.storage.journal.verbosity setting.

STORAGE is the parent component of JOURNAL. If systemLog.component.storage.journal.verbosity is unset, MongoDB uses the STORAGE verbosity level for JOURNAL components.

NETWORK

Messages related to network activities, such as accepting connections. To specify the log level for NETWORK components, set the systemLog.component.network.verbosity parameter.

QUERY

Messages related to queries, including query planner activities. To specify the log level for QUERY components, set the systemLog.component.query.verbosity parameter.

QUERYSTATS

Messages related to $queryStats operations. To specify the log level for QUERYSTATS components, set the systemLog.component.queryStats.verbosity parameter.

RECOVERY

Messages related to storage recovery activities. To specify the log level for RECOVERY components, use the systemLog.component.storage.recovery.verbosity setting.

STORAGE is the parent component of RECOVERY. If systemLog.component.storage.recovery.verbosity is unset, MongoDB uses the STORAGE verbosity level for RECOVERY components.

REPL

Messages related to replica sets, such as initial sync, heartbeats, steady state replication, and rollback. [2] To specify the log level for REPL components, set the systemLog.component.replication.verbosity parameter.

REPL is the parent component of the ELECTION, INITSYNC, REPL_HB, and ROLLBACK components.

REPL_HB

Messages related specifically to replica set heartbeats. To specify the log level for REPL_HB components, set the systemLog.component.replication.heartbeats.verbosity parameter.

REPL is the parent component of REPL_HB. If systemLog.component.replication.heartbeats.verbosity is unset, MongoDB uses the REPL verbosity level for REPL_HB components.

ROLLBACK

Messages related to rollback operations. To specify the log level for ROLLBACK components, set the systemLog.component.replication.rollback.verbosity parameter.

REPL is the parent component of ROLLBACK. If systemLog.component.replication.rollback.verbosity is unset, MongoDB uses the REPL verbosity level for ROLLBACK components.

SHARDING

Messages related to sharding activities, such as the startup of the mongos. To specify the log level for SHARDING components, use the systemLog.component.sharding.verbosity setting.

STORAGE

Messages related to storage activities, such as processes involved in the fsync command. To specify the log level for STORAGE components, use the systemLog.component.storage.verbosity setting.

STORAGE is the parent component of JOURNAL and RECOVERY.

TXN

Messages related to multi-document transactions. To specify the log level for TXN components, use the systemLog.component.transaction.verbosity setting.

WRITE

Messages related to write operations, such as update commands. To specify the log level for WRITE components, use the systemLog.component.write.verbosity setting.

WT

New in version 5.3.

Messages related to the WiredTiger storage engine. To specify the log level for WT components, use the systemLog.component.storage.wt.verbosity setting.

WTBACKUP

New in version 5.3.

Messages related to backup operations performed by the WiredTiger storage engine. To specify the log level for the WTBACKUP components, use the systemLog.component.storage.wt.wtBackup.verbosity setting.

WTCHKPT

New in version 5.3.

Messages related to checkpoint operations performed by the WiredTiger storage engine. To specify the log level for WTCHKPT components, use the systemLog.component.storage.wt.wtCheckpoint.verbosity setting.

WTCMPCT

New in version 5.3.

Messages related to compaction operations performed by the WiredTiger storage engine. To specify the log level for WTCMPCT components, use the systemLog.component.storage.wt.wtCompact.verbosity setting.

WTEVICT

New in version 5.3.

Messages related to eviction operations performed by the WiredTiger storage engine. To specify the log level for WTEVICT components, use the systemLog.component.storage.wt.wtEviction.verbosity setting.

WTHS

New in version 5.3.

Messages related to the history store of the WiredTiger storage engine. To specify the log level for WTHS components, use the systemLog.component.storage.wt.wtHS.verbosity setting.

WTRECOV

New in version 5.3.

Messages related to recovery operations performed by the WiredTiger storage engine. To specify the log level for WTRECOV components, use the systemLog.component.storage.wt.wtRecovery.verbosity setting.

WTRTS

New in version 5.3.

Messages related to rollback to stable (RTS) operations performed by the WiredTiger storage engine. To specify the log level for WTRTS components, use the systemLog.component.storage.wt.wtRTS.verbosity setting.

WTSLVG

New in version 5.3.

Messages related to salvage operations performed by the WiredTiger storage engine. To specify the log level for WTSLVG components, use the systemLog.component.storage.wt.wtSalvage.verbosity setting.

WTTS

New in version 5.3.

Messages related to timestamps used by the WiredTiger storage engine. To specify the log level for WTTS components, use the systemLog.component.storage.wt.wtTimestamp.verbosity setting.

WTTXN

New in version 5.3.

Messages related to transactions performed by the WiredTiger storage engine. To specify the log level for WTTXN components, use the systemLog.component.storage.wt.wtTransaction.verbosity setting.

WTVRFY

New in version 5.3.

Messages related to verification operations performed by the WiredTiger storage engine. To specify the log level for WTVRFY components, use the systemLog.component.storage.wt.wtVerify.verbosity setting.

WTWRTLOG

New in version 5.3.

Messages related to log write operations performed by the WiredTiger storage engine. To specify the log level for WTWRTLOG components, use the systemLog.component.storage.wt.wtWriteLog.verbosity setting.

-

Messages not associated with a named component. Unnamed components have the default log level specified in the systemLog.verbosity setting. The systemLog.verbosity setting is the default setting for both named and unnamed components.

See Filtering by Component for log parsing examples that filter on the component field.

MongoDB Drivers and client applications (including mongosh) have the ability to send identifying information at the time of connection to the server. After the connection is established, the client does not send the identifying information again unless the connection is dropped and reestablished.

This identifying information is contained in the attributes field of the log entry. The exact information included varies by client.

Below is a sample log message containing the client data document as transmitted from a mongosh connection. The client data is contained in the doc object in the attributes field:

{"t":{"$date":"2020-05-20T16:21:31.561+00:00"},"s":"I", "c":"NETWORK", "id":51800, "ctx":"conn202", "svc": "R", "msg":"client metadata", "attr":{"remote":"127.0.0.1:37106","client":"conn202","doc":{"application":{"name":"MongoDB Shell"},"driver":{"name":"MongoDB Internal Client","version":"4.4.0"},"os":{"type":"Linux","name":"CentOS Linux release 8.0.1905 (Core) ","architecture":"x86_64","version":"Kernel 4.18.0-80.11.2.el8_0.x86_64"}}}}

When secondary members of a replica set initiate a connection to a primary, they send similar data. A sample log message containing this initiation connection might appear as follows. The client data is contained in the doc object in the attributes field:

{"t":{"$date":"2020-05-20T16:33:40.595+00:00"},"s":"I", "c":"NETWORK", "id":51800, "ctx":"conn214", "svc": "R", "msg":"client metadata", "attr":{"remote":"127.0.0.1:37176","client":"conn214","doc":{"driver":{"name":"NetworkInterfaceTL","version":"4.4.0"},"os":{"type":"Linux","name":"CentOS Linux release 8.0.1905 (Core) ","architecture":"x86_64","version":"Kernel 4.18.0-80.11.2.el8_0.x86_64"}}}}

See the examples section for a pretty-printed example showing client data.

For a complete description of client information and required fields, see the MongoDB Handshake specification.

You can specify the logging verbosity level to increase or decrease the amount of log messages MongoDB outputs. Verbosity levels can be adjusted for all components together, or for specific named components individually.

Verbosity affects log entries in the severity categories Informational and Debug only. Severity categories above these levels are always shown.

You might set verbosity levels to a high value to show detailed logging for debugging or development, or to a low value to minimize writes to the log on a vetted production deployment. [2]

To view the current verbosity levels, use the db.getLogComponents() method:

db.getLogComponents()

Your output might resemble the following:

{
"verbosity" : 0,
"accessControl" : {
"verbosity" : -1
},
"command" : {
"verbosity" : -1
},
...
"storage" : {
"verbosity" : -1,
"recovery" : {
"verbosity" : -1
},
"journal" : {
"verbosity" : -1
}
},
...

The initial verbosity entry is the parent verbosity level for all components, while the individual named components that follow, such as accessControl, indicate the specific verbosity level for that component, overriding the global verbosity level for that particular component if set.

A value of -1, indicates that the component inherits the verbosity level of their parent, if they have one (as with recovery above, inheriting from storage), or the global verbosity level if they do not (as with command). Inheritance relationships for verbosity levels are indicated in the components section.

You can configure the verbosity level using: the systemLog.verbosity and systemLog.component.<name>.verbosity settings, the logComponentVerbosity parameter, or the db.setLogLevel() method. [2]

To configure the default log level for all components, use the systemLog.verbosity setting. To configure the level of specific components, use the systemLog.component.<name>.verbosity settings.

For example, the following configuration sets the systemLog.verbosity to 1, the systemLog.component.query.verbosity to 2, the systemLog.component.storage.verbosity to 2, and the systemLog.component.storage.journal.verbosity to 1:

systemLog:
verbosity: 1
component:
query:
verbosity: 2
storage:
verbosity: 2
journal:
verbosity: 1

You would set these values in the configuration file or on the command line for your mongod or mongos instance.

All components not specified explicitly in the configuration have a verbosity level of -1, indicating that they inherit the verbosity level of their parent, if they have one, or the global verbosity level (systemLog.verbosity) if they do not.

To set the logComponentVerbosity parameter, pass a document with the verbosity settings to change.

For example, the following sets the default verbosity level to 1, the query to 2, the storage to 2, and the storage.journal to 1.

db.adminCommand( {
setParameter: 1,
logComponentVerbosity: {
verbosity: 1,
query: {
verbosity: 2
},
storage: {
verbosity: 2,
journal: {
verbosity: 1
}
}
}
} )

You would set these values from mongosh.

Use the db.setLogLevel() method to update a single component log level. For a component, you can specify verbosity level of 0 to 5, or you can specify -1 to inherit the verbosity of the parent. For example, the following sets the systemLog.component.query.verbosity to its parent verbosity (i.e. default verbosity):

db.setLogLevel(-1, "query")

You would set this value from mongosh.

[2](1, 2, 3, 4, 5) Secondary members of a replica set now log oplog entries that take longer than the slow operation threshold to apply. These slow oplog messages:
  • Are logged for the secondaries in the diagnostic log.
  • Are logged under the REPL component with the text applied op: <oplog entry> took <num>ms.
  • Do not depend on the log levels (either at the system or component level)
  • Do not depend on the profiling level.
  • Are affected by slowOpSampleRate.
The profiler does not capture slow oplog entries.

Client operations (such as queries) appear in the log if their duration exceeds the slow operation threshold or when the log verbosity level is 1 or higher. [2] These log entries include the full command object associated with the operation.

The profiler entries and the diagnostic log messages (i.e. mongod/mongos logmessages) for read/write operations include:

  • planCacheShapeHash to help identify slow queries with the same plan cache query shape.

    Starting in MongoDB 8.0, the pre-existing queryHash field is renamed to planCacheShapeHash. If you're using an earlier MongoDB version, you'll see queryHash instead of planCacheShapeHash.

  • planCacheKey to provide more insight into the query plan cache for slow queries.

Starting in MongoDB 5.0, slow operation log messages include a remote field specifying client IP address.

Starting in MongoDB 6.2, slow operation log messages include a queryFramework field that indicates which query engine executed the query:

  • queryFramework: "classic" indicates that the classic engine executed the query.

  • queryFramework: "sbe" indicates that the slot-based query execution engine executed the query.

Starting in MongoDB 6.1, slow operation log messages include cache refresh time fields.

Starting in MongoDB 6.3, slow operation log messages and database profiler entries include a cpuNanos field that specifies the total CPU time spent by a query operation in nanoseconds. The cpuNanos field is only available on Linux systems.

Starting in MongoDB 7.0 (and 6.0.13, 5.0.24), the totalOplogSlotDurationMicros in the slow query log message shows the time between a write operation getting a commit timestamp to commit the storage engine writes and actually committing. mongod supports parallel writes. However, it commits write operations with commit timestamps in any order.

Example

Consider the following writes with commit timestamps:

  • writeA with Timestamp1

  • writeB with Timestamp2

  • writeC with Timestamp3

Suppose writeB commits first at Timestamp2. Replication is paused until writeA commits because writeA's oplog entry with Timestamp1 is required for replication to copy the oplog to secondary replica set members.

For a pretty-printed example of a slow operation log entry, see Log Message Examples.

Starting in MongoDB 8.0, the queryShapeHash field for a query shape is also included in the slow query log when available.

Also starting in MongoDB 8.0, the slow query output includes a queues document that contains information about the operation's queues. Each queue in the queues field contains a totalTimeQueuedMicros field that contains the total cumulative time, in microseconds, that the operation spent in the corresponding queue.

New in version 5.0.

Starting in MongoDB 5.0, you can use the remoteOpWaitMillis log field to obtain the wait time (in milliseconds) for results from shards.

remoteOpWaitMillis is only logged:

To determine if a merge operation or a shard issue is causing a slow query, compare the workingMillis and remoteOpWaitMillis time fields in the log. workingMillis is the total time the query took to complete. Specifically:

  • If workingMillis is slightly longer than remoteOpWaitMillis, then most of the time was spent waiting for a shard response. For example, workingMillis of 17 and remoteOpWaitMillis of 15.

  • If workingMillis is significantly longer than remoteOpWaitMillis, then most of the time was spent performing the merge. For example, workingMillis of 100 and remoteOpWaitMillis of 15.

When using Queryable Encryption, CRUD operations against encrypted collections are omitted from the slow query log. For details, see Queryable Encryption redaction.

Available in MongoDB Enterprise only

A mongod or mongos running with redactClientLogData redacts any message accompanying a given log event before logging, leaving only metadata, source files, or line numbers related to the event. redactClientLogData prevents potentially sensitive information from entering the system log at the cost of diagnostic detail.

For example, the following operation inserts a document into a mongod running without log redaction. The mongod has the log verbosity level set to 1:

db.clients.insertOne( { "name" : "Joe", "PII" : "Sensitive Information" } )

This operation produces the following log event:

{
"t": { "$date": "2024-07-19T15:36:55.024-07:00" },
"s": "I",
"c": "COMMAND",
...
"attr": {
"type": "command",
...
"appName": "mongosh 2.2.10",
"command": {
"insert": "clients",
"documents": [
{
"name": "Joe",
"PII": "Sensitive Information",
"_id": { "$oid": "669aea8792c7fd822d3e1d8c" }
}
],
"ordered": true,
...
}
...
}
}

When mongod runs with redactClientLogData and performs the same insert operation, it produces the following log event:

{
"t": { "$date": "2024-07-19T15:36:55.024-07:00" },
"s": "I",
"c": "COMMAND",
...
"attr": {
"type": "command",
...
"appName": "mongosh 2.2.10",
"command": {
"insert": "###",
"documents": [
{
"name": "###",
"PII": "###",
"_id": "###"
}
],
"ordered": "###",
...
}
...
}
}

Use redactClientLogData in conjunction with Encryption at Rest and TLS/SSL (Transport Encryption) to assist compliance with regulatory requirements.

Log parsing is the act of programmatically searching through and analyzing log files, often in an automated manner. With the introduction of structured logging, log parsing is made simpler and more powerful. For example:

  • Log message fields are presented as key-value pairs. Log parsers can query by specific keys of interest to efficiently filter results.

  • Log messages always contain the same message structure. Log parsers can reliably extract information from any log message, without needing to code for cases where information is missing or formatted differently.

The following examples demonstrate common log parsing workflows when working with MongoDB JSON log output.

When working with MongoDB structured logging, the third-party jq command-line utility is a useful tool that allows for easy pretty-printing of log entries, and powerful key-based matching and filtering.

jq is an open-source JSON parser, and is available for Linux, Windows, and macOS.

These examples use jq to simplify log parsing.

The following example shows the top 10 unique message values in a given log file, sorted by frequency:

jq -r ".msg" /var/log/mongodb/mongod.log | sort | uniq -c | sort -rn | head -10

Remote client connections are shown in the log under the "remote" key in the attribute object. The following counts all unique connections over the course of the log file and presents them in descending order by number of occurrences:

jq -r '.attr.remote' /var/log/mongodb/mongod.log | grep -v 'null' | sort | uniq -c | sort -r

Note that connections from the same IP address, but connecting over different ports, are treated as different connections by this command. You could limit output to consider IP addresses only, with the following change:

jq -r '.attr.remote' /var/log/mongodb/mongod.log | grep -v 'null' | awk -F':' '{print $1}' | sort | uniq -c | sort -r

The following example counts all remote MongoDB driver connections, and presents each driver type and version in descending order by number of occurrences:

jq -cr '.attr.doc.driver' /var/log/mongodb/mongod.log | grep -v null | sort | uniq -c | sort -rn

The following example analyzes the reported client data of remote MongoDB driver connections and client applications, including mongosh, and prints a total for each unique operating system type that connected, sorted by frequency:

jq -r '.attr.doc.os.type' /var/log/mongodb/mongod.log | grep -v null | sort | uniq -c | sort -rn

The string "Darwin", as reported in this log field, represents a macOS client.

With slow operation logging enabled, the following returns only the slow operations that took above 2000 milliseconds:, for further analysis:

jq 'select(.attr.workingMillis>=2000)' /var/log/mongodb/mongod.log

Consult the jq documentation for more information on the jq filters shown in this example.

Log components (the third field in the JSON log output format) indicate the general category a given log message falls under. Filtering by component is often a great starting place when parsing log messages for relevant events.

The following example prints only the log messages of component type REPL:

jq 'select(.c=="REPL")' /var/log/mongodb/mongod.log

The following example prints all log messages except those of component type REPL:

jq 'select(.c!="REPL")' /var/log/mongodb/mongod.log

The following example print log messages of component type REPL or STORAGE:

jq 'select( .c as $c | ["REPL", "STORAGE"] | index($c) )' /var/log/mongodb/mongod.log

Consult the jq documentation for more information on the jq filters shown in this example.

Log IDs (the fifth field in the JSON log output format) map to specific log events, and can be relied upon to remain stable over successive MongoDB releases.

As an example, you might be interested in the following two log events, showing a client connection followed by a disconnection:

{"t":{"$date":"2020-06-01T13:06:59.027-0500"},"s":"I", "c":"NETWORK", "id":22943, "ctx":"listener", "svc": "R", "msg":"connection accepted from {session_remote} #{session_id} ({connectionCount}{word} now open)", "attr":{"session_remote":"127.0.0.1:61298", "session_id":164,"connectionCount":11,"word":" connections"}}
{"t":{"$date":"2020-06-01T13:07:03.490-0500"},"s":"I", "c":"NETWORK", "id":22944, "ctx":"conn157", "svc": "R", "msg":"end connection {remote} ({connectionCount}{word} now open)", "attr":{"remote":"127.0.0.1:61298","connectionCount":10,"word":" connections"}}

The log IDs for these two entries are 22943 and 22944 respectively. You could then filter your log output to show only these log IDs, effectively showing only client connection activity, using the following jq syntax:

jq 'select( .id as $id | [22943, 22944] | index($id) )' /var/log/mongodb/mongod.log

Consult the jq documentation for more information on the jq filters shown in this example.

Log output can be further refined by filtering on the timestamp field, limiting log entries returned to a specific date range. For example, the following returns all log entries that occurred on April 15th, 2020:

jq 'select(.t["$date"] >= "2020-04-15T00:00:00.000" and .t["$date"] <= "2020-04-15T23:59:59.999")' /var/log/mongodb/mongod.log

Note that this syntax includes the full timestamp, including milliseconds but excluding the timezone offset.

Filtering by date range can be combined with any of the examples above, creating weekly reports or yearly summaries for example. The following syntax expands the "Monitoring Connections" example from earlier to limit results to the month of May, 2020:

jq 'select(.t["$date"] >= "2020-05-01T00:00:00.000" and .t["$date"] <= "2020-05-31T23:59:59.999" and .attr.remote)' /var/log/mongodb/mongod.log

Consult the jq documentation for more information on the jq filters shown in this example.

Log ingestion services are third-party products that intake and aggregate log files, usually from a distributed cluster of systems, and provide ongoing analysis of that data in a central location.

The JSON log format allows for more flexibility when working with log ingestion and analysis services. Whereas plaintext logs generally require some manner of transformation before being eligible for use with these products, JSON files can often be consumed out of the box, depending on the service. Further, JSON-formatted logs offer more control when performing filtering for these services, as the key-value structure offers the ability to specifically import only the fields of interest, while omitting the rest.

Consult the documentation for your chosen third-party log ingestion service for more information.

The following examples show log messages in JSON output format.

These log messages are presented in pretty-printed format for convenience.

This example shows a startup warning:

{
"t": {
"$date": "2020-05-20T19:17:06.188+00:00"
},
"s": "W",
"c": "CONTROL",
"id": 22120,
"ctx": "initandlisten",
"svc": "R",
"msg": "Access control is not enabled for the database. Read and write access to data and configuration is unrestricted",
"tags": [
"startupWarnings"
]
}

This example shows a client connection that includes client data:

{
"t": {
"$date": "2020-05-20T19:18:40.604+00:00"
},
"s": "I",
"c": "NETWORK",
"id": 51800,
"ctx": "conn281",
"svc": "R",
"msg": "client metadata",
"attr": {
"remote": "192.168.14.15:37666",
"client": "conn281",
"doc": {
"application": {
"name": "MongoDB Shell"
},
"driver": {
"name": "MongoDB Internal Client",
"version": "4.4.0"
},
"os": {
"type": "Linux",
"name": "CentOS Linux release 8.0.1905 (Core) ",
"architecture": "x86_64",
"version": "Kernel 4.18.0-80.11.2.el8_0.x86_64"
}
}
}
}

Starting in MongoDB 8.0, slow operations are logged based on the time that MongoDB spends working on that operation, rather than the total latency for the operation.

You can use the metrics in the slow operation log to identify where an operation spends time in its lifecycle, which helps identify possible performance improvements.

In the following example log message:

  • The amount of time spent waiting for resources while executing the query is shown in these metrics:

    • queues.execution.totalTimeQueuedMicros

    • timeAcquiringMicros

  • workingMillis is the amount of time that MongoDB spends working on the operation.

  • durationMillis is the operation's total latency.

{
"t":{
"$date":"2024-06-01T13:24:10.034+00:00"
},
"s":"I",
"c":"COMMAND",
"id":51803,
"ctx":"conn3",
"msg":"Slow query",
"attr":{
"type":"command",
"isFromUserConnection":true,
"ns":"db.coll",
"collectionType":"normal",
"appName":"MongoDB Shell",
"command":{
"find":"coll",
"filter":{
"b":-1
},
"sort":{
"splitPoint":1
},
"readConcern":{ },
"$db":"db"
},
"planSummary":"COLLSCAN",
"planningTimeMicros":87,
"keysExamined":0,
"docsExamined":20889,
"hasSortStage":true,
"nBatches":1,
"cursorExhausted":true,
"numYields":164,
"nreturned":99,
"planCacheShapeHash":"9C05019A",
"planCacheKey":"C41063D6",
"queryFramework":"classic",
"reslen":96,
"locks":{
"ReplicationStateTransition":{
"acquireCount":{
"w":3
}
},
"Global":{
"acquireCount":{
"r":327,
"w":1
}
},
"Database":{
"acquireCount":{
"r":1
},
"acquireWaitCount":{
"r":1
},
"timeAcquiringMicros":{
"r":2814
}
},
"Collection":{
"acquireCount":{
"w":1
}
}
},
"flowControl":{
"acquireCount":1,
"acquireWaitCount":1,
"timeAcquiringMicros":8387
},
"readConcern":{
"level":"local",
"provenance":"implicitDefault"
},
"storage":{ },
"cpuNanos":20987385,
"remote":"127.0.0.1:47150",
"protocol":"op_msg",
"queues":{
"ingress":{
"admissions":7,
"totalTimeQueuedMicros":0
},
"execution":{
"admissions":328,
"totalTimeQueuedMicros":2109
}
},
"workingMillis":89,
"durationMillis":101
}
}

Starting in MongoDB 8.0, the pre-existing queryHash field is renamed to planCacheShapeHash. If you're using an earlier MongoDB version, you'll see queryHash instead of planCacheShapeHash.

This example demonstrates character escaping, as shown in the setName field of the attribute object:

{
"t": {
"$date": "2020-05-20T19:11:09.268+00:00"
},
"s": "I",
"c": "REPL",
"id": 21752,
"ctx": "ReplCoord-0",
"svc": "R",
"msg": "Scheduling remote command request",
"attr": {
"context": "vote request",
"request": "RemoteCommand 229 -- target:localhost:27003 db:admin cmd:{ replSetRequestVotes: 1, setName: \"my-replica-name\", dryRun: true, term: 3, candidateIndex: 0, configVersion: 2, configTerm: 3, lastAppliedOpTime: { ts: Timestamp(1589915409, 1), t: 3 } }"
}
}

Starting in MongoDB 5.0, log messages for slow queries on views include a resolvedViews field that contains the view details:

"resolvedViews": [ {
"viewNamespace": <String>, // namespace and view name
"dependencyChain": <Array of strings>, // view name and collection
"resolvedPipeline": <Array of documents> // aggregation pipeline for view
} ]

The following example uses the test database and creates a view named myView that sorts the documents in myCollection by the firstName field:

use test
db.createView( "myView", "myCollection", [ { $sort: { "firstName" : 1 } } ] )

Assume a slow query is run on myView. The following example log message contains a resolvedViews field for myView:

{
"t": {
"$date": "2021-09-30T17:53:54.646+00:00"
},
"s": "I",
"c": "COMMAND",
"id": 51803,
"ctx": "conn249",
"svc": "R",
"msg": "Slow query",
"attr": {
"type": "command",
"ns": "test.myView",
"appName": "MongoDB Shell",
"command": {
"find": "myView",
"filter": {},
"lsid": {
"id": { "$uuid": "ad176471-60e5-4e82-b977-156a9970d30f" }
},
"$db": "test"
},
"planSummary":"COLLSCAN",
"resolvedViews": [ {
"viewNamespace": "test.myView",
"dependencyChain": [ "myView", "myCollection" ],
"resolvedPipeline": [ { "$sort": { "firstName": 1 } } ]
} ],
"keysExamined": 0,
"docsExamined": 1,
"hasSortStage": true,
"cursorExhausted": true,
"numYields": 0,
"nreturned": 1,
"planCacheShapeHash": "3344645B",
"planCacheKey": "1D3DE690",
"queryFramework": "classic"
"reslen": 134,
"locks": { "ParallelBatchWriterMode": { "acquireCount": { "r": 1 } },
"ReplicationStateTransition": { "acquireCount": { "w": 1 } },
"Global": { "acquireCount": { "r": 4 } },
"Database": { "acquireCount": {"r": 1 } },
"Collection": { "acquireCount": { "r": 1 } },
"Mutex": { "acquireCount": { "r": 4 } } },
"storage": {},
"remote": "127.0.0.1:34868",
"protocol": "op_msg",
"workingMillis": 0,
"durationMillis": 0
}
}
}

Starting in MongoDB 8.0, the pre-existing queryHash field is renamed to planCacheShapeHash. If you're using an earlier MongoDB version, you'll see queryHash instead of planCacheShapeHash.

Starting in MongoDB 5.0, log messages for slow queries include a system.profile.authorization section. These metrics help determine if a request is delayed because of contention for the user authorization cache.

"authorization": {
"startedUserCacheAcquisitionAttempts": 1,
"completedUserCacheAcquisitionAttempts": 1,
"userCacheWaitTimeMicros": 508
},

Starting in MongoDB 6.3, a message is added to the log if the time to send an operation response exceeds the slowms threshold option.

The message is known as a session workflow log message and contains various times to perform an operation in a database session.

Example session workflow log message:

{
"t": {
"$date": "2022-12-14T17:22:44.233+00:00"
},
"s": "I",
"c": "EXECUTOR",
"id": 6983000,
"ctx": "conn1",
"svc": "R",
"msg": "Slow network response send time",
"attr": {
"elapsed": {
"totalMillis": 109,
"activeMillis": 30,
"receiveWorkMillis": 2,
"processWorkMillis": 10,
"sendResponseMillis": 22,
"yieldMillis": 15,
"finalizeMillis": 30
}
}
}

The times are in milliseconds.

A session workflow message is added to the log if sendResponseMillis exceeds the slowms threshold option.

Field
Description
totalMillis
Total time to perform the operation in the session, which includes the time spent waiting for a message to be received.
activeMillis
Time between receiving a message and completing the operation associated with that message. Time includes sending a response and performing any clean up.
receivedWorkMillis
Time to receive the operation information over the network.
processWorkMillis
Time to process the operation and create the response.
sendResponseMillis
Time to send the response.
yieldMillis
Time between releasing the worker thread and the thread being used again.
finalize
Time to end and close the session workflow.

Starting in MongoDB 6.3, a message is added to the log if the time that an operation waited between acquisition of a server connection and writing the bytes to send to the server over the network exceeds 1 millisecond.

By default, the message is logged at the "I" information level, and at most once every second to avoid too many log messages. If you must obtain every log message, change your log level to debug.

If the operation wait time exceeds 1 millisecond and the message is logged at the information level within the last second, then the next message is logged at the debug level. Otherwise, the next message is logged at the information level.

Example log message:

{
"t": {
"$date":"2023-01-31T15:22:29.473+00:00"
},
"s": "I",
"c": "NETWORK",
"id": 6496702,
"ctx": "ReplicaSetMonitor-TaskExecutor",
"svc": "R",
"msg": "Acquired connection for remote operation and completed writing to wire",
"attr": {
"durationMicros": 1683
}
}

The following table describes the durationMicros field in attr.

Field
Description
durationMicros
Time in microseconds that the operation waited between acquisition of a server connection and writing the bytes to send to the server over the network.

Starting in MongoDB 6.1, log messages for slow queries include the following cache refresh time fields:

  • catalogCacheDatabaseLookupDurationMillis

  • catalogCacheCollectionLookupDurationMillis

  • databaseVersionRefreshDurationMillis

  • shardVersionRefreshMillis

Starting in MongoDB 7.0, log messages for slow queries also include the catalogCacheIndexLookupDurationMillis field that indicates the time that the operation spent fetching information from the index cache. This release also renames the shardVersionRefreshMillis field to placementVersionRefreshMillis.

The following example includes:

  • catalogCacheDatabaseLookupDurationMillis

  • catalogCacheCollectionLookupDurationMillis

  • catalogCacheIndexLookupDurationMillis

{
"t": {
"$date": "2023-03-17T09:47:55.929+00:00"
},
"s": "I",
"c": "COMMAND",
"id": 51803,
"ctx": "conn14",
"svc": "R",
"msg": "Slow query",
"attr": {
"type": "command",
"ns": "db.coll",
"appName": "MongoDB Shell",
"command": {
"insert": "coll",
"ordered": true,
"lsid": {
"id": {
"$uuid": "5d50b19c-8559-420a-a122-8834e012274a"
}
},
"$clusterTime": {
"clusterTime": {
"$timestamp": {
"t": 1679046398,
"i": 8
}
},
"signature": {
"hash": {
"$binary": {
"base64": "AAAAAAAAAAAAAAAAAAAAAAAAAAA=",
"subType": "0"
}
},
"keyId": 0
}
},
"$db": "db"
},
"catalogCacheDatabaseLookupDurationMillis": 19,
"catalogCacheCollectionLookupDurationMillis": 68,
"catalogCacheIndexLookupDurationMillis": 16026,
"nShards": 1,
"ninserted": 1,
"numYields": 232,
"reslen": 96,
"readConcern": {
"level": "local",
"provenance": "implicitDefault",
},
"cpuNanos": 29640339,
"remote": "127.0.0.1:48510",
"protocol": "op_msg",
"remoteOpWaitMillis": 4078,
"workingMillis": 20334,
"durationMillis": 20334
}
}

In a Linux system, messages are subject to the rules defined in the Linux configuration file /etc/systemd/journald.conf. By default, log message bursts are limited to 1000 messages within a 30 second period. To see more messages, increase the RateLimitBurst parameter in /etc/systemd/journald.conf.

You can use MongoDB Atlas to download a zipped file containing the logs for a selected hostname or process in your database deployment. To learn more, see View and Download MongoDB Logs.

Back

Glossary