Menu

Caution

Grafana Alloy is the new name for our distribution of the OTel collector. Grafana Agent has been deprecated and is in Long-Term Support (LTS) through October 31, 2025. Grafana Agent will reach an End-of-Life (EOL) on November 1, 2025. Read more about why we recommend migrating to Grafana Alloy.

Important: This documentation is about an older version. It's relevant only to the release noted, many of the features and functions have been updated or replaced. Please view the current version.

Open source

loki.process

loki.process receives log entries from other loki components, applies one or more processing stages, and forwards the results to the list of receivers in the component’s arguments.

A stage is a multi-purpose tool that can parse, transform, and filter log entries before they’re passed to a downstream component. These stages are applied to each log entry in order of their appearance in the configuration file. All stages within a loki.process block have access to the log entry’s label set, the log line, the log timestamp, as well as a shared map of ’extracted’ values so that the results of one stage can be used in a subsequent one.

Multiple loki.process components can be specified by giving them different labels.

Usage

river
loki.process "LABEL" {
  forward_to = RECEIVER_LIST

  stage.STAGENAME {
    ...
  }
  ...
}

Arguments

loki.process supports the following arguments:

NameTypeDescriptionDefaultRequired
forward_tolist(LogsReceiver)Where to forward log entries after processing.yes

Blocks

The following blocks are supported inside the definition of loki.process:

HierarchyBlockDescriptionRequired
stage.cristage.criConfigures a pre-defined CRI-format pipeline.no
stage.dockerstage.dockerConfigures a pre-defined Docker log format pipeline.no
stage.dropstage.dropConfigures a drop processing stage.no
stage.jsonstage.jsonConfigures a JSON processing stage.no
stage.label_dropstage.label_dropConfigures a label_drop processing stage.no
stage.label_keepstage.label_keepConfigures a label_keep processing stage.no
stage.labelsstage.labelsConfigures a labels processing stage.no
stage.structured_metadatastage.structured_metadataConfigures a structured metadata processing stage.no
stage.limitstage.limitConfigures a limit processing stage.no
stage.logfmtstage.logfmtConfigures a logfmt processing stage.no
stage.matchstage.matchConfigures a match processing stage.no
stage.metricsstage.metricsConfigures a metrics stage.no
stage.multilinestage.multilineConfigures a multiline processing stage.no
stage.outputstage.outputConfigures an output processing stage.no
stage.packstage.packConfigures a pack processing stage.no
stage.regexstage.regexConfigures a regex processing stage.no
stage.replacestage.replaceConfigures a replace processing stage.no
stage.static_labelsstage.static_labelsConfigures a static_labels processing stage.no
stage.templatestage.templateConfigures a template processing stage.no
stage.tenantstage.tenantConfigures a tenant processing stage.no
stage.timestampstage.timestampConfigures a timestamp processing stage.no
stage.geoipstage.geoipConfigures a geoip processing stage.no

A user can provide any number of these stage blocks nested inside loki.process; these will run in order of appearance in the configuration file.

stage.cri block

The stage.cri inner block enables a predefined pipeline which reads log lines using the CRI logging format.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
max_partial_linesnumberMaximum number of partial lines to hold in memory.100no
max_partial_line_sizenumberMaximum number of characters which a partial line can have.0no
max_partial_line_size_truncateboolTruncate partial lines that are longer than max_partial_line_size.falseno

max_partial_line_size is only taken into account if max_partial_line_size_truncate is set to true.

river
stage.cri {}

CRI specifies log lines as single space-delimited values with the following components:

  • time: The timestamp string of the log
  • stream: Either stdout or stderr
  • flags: CRI flags including F or P
  • log: The contents of the log line

Given the following log line, the subsequent key-value pairs are created in the shared map of extracted data:

"2019-04-30T02:12:41.8443515Z stdout F message"

content: message
stream: stdout
timestamp: 2019-04-30T02:12:41.8443515

stage.docker block

The stage.docker inner block enables a predefined pipeline which reads log lines in the standard format of Docker log files.

The stage.docker block does not support any arguments or inner blocks, so it is always empty.

river
stage.docker {}

Docker log entries are formatted as JSON with the following keys:

  • log: The content of log line
  • stream: Either stdout or stderr
  • time: The timestamp string of the log line

Given the following log line, the subsequent key-value pairs are created in the shared map of extracted data:

{"log":"log message\n","stream":"stderr","time":"2019-04-30T02:12:41.8443515Z"}

output: log message\n
stream: stderr
timestamp: 2019-04-30T02:12:41.8443515

stage.drop block

The stage.drop inner block configures a filtering stage that drops log entries based on several options. If multiple options are provided, they’re treated as AND clauses and must all be true for the log entry to be dropped. To drop entries with an OR clause, specify multiple drop blocks in sequence.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
sourcestringName from extracted data to parse. If empty or not defined, it uses the log message.""no
expressionstringA valid RE2 regular expression.""no
valuestringIf both source and value are specified, the stage drops lines where value exactly matches the source content.""no
older_thandurationIf specified, the stage drops lines whose timestamp is older than the current time minus this duration.""no
longer_thanstringIf specified, the stage drops lines whose size exceeds the configured value.""no
drop_counter_reasonstringA custom reason to report for dropped lines."drop_stage"no

The expression field needs to be a RE2 regex string. If source is empty or not provided, the regex attempts to match the log line itself. If source is provided, the regex attempts to match the corresponding value from the extracted map.

The value field can only work with values from the extracted map, and must be specified together with source. Entries are dropped when there is an exact match between the two.

Whenever an entry is dropped, the metric loki_process_dropped_lines_total is incremented. By default, the reason label is "drop_stage", but you can provide a custom label using the drop_counter_reason argument.

The following stage drops log entries that contain the word debug and are longer than 1KB.

river
stage.drop {
    expression  = ".*debug.*"
    longer_than = "1KB"
}

On the following example, we define multiple drop blocks so loki.process drops entries that are either 24h or older, are longer than 8KB, or the extracted value of ‘app’ is equal to foo.

river
stage.drop {
    older_than          = "24h"
    drop_counter_reason = "too old"
}

stage.drop {
    longer_than         = "8KB"
    drop_counter_reason = "too long"
}

stage.drop {
    source = "app"
    value  = "foo"
}

stage.json block

The stage.json inner block configures a JSON processing stage that parses incoming log lines or previously extracted values as JSON and uses JMESPath expressions to extract new values from them.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
expressionsmap(string)Key-value pairs of JMESPath expressions.yes
sourcestringSource of the data to parse as JSON.""no
drop_malformedboolDrop lines whose input cannot be parsed as valid JSON.falseno

When configuring a JSON stage, the source field defines the source of data to parse as JSON. By default, this is the log line itself, but it can also be a previously extracted value.

The expressions field is the set of key-value pairs of JMESPath expressions to run. The map key defines the name with which the data is extracted, while the map value is the expression used to populate the value.

Here’s a given log line and two JSON stages to run.

river
{"log":"log message\n","extra":"{\"user\":\"agent\"}"}

loki.process "username" {
  stage.json {
      expressions = {output = "log", extra = ""}
  }

  stage.json {
      source      = "extra"
      expressions = {username = "user"}
  }
}

In this example, the first stage uses the log line as the source and populates these values in the shared map. An empty expression means using the same value as the key (so extra="extra").

output: log message\n
extra: {"user": "agent"}

The second stage uses the value in extra as the input and appends the following key-value pair to the set of extracted data.

username: agent

Note

Due to a limitation of the upstream jmespath library, you must wrap any string that contains a hyphen - in quotes so that it’s not considered a numerical expression.

If you don’t use quotes to wrap a string that contains a hyphen, you will get errors like: Unexpected token at the end of the expression: tNumber

You can use one of two options to circumvent this issue:

  1. An escaped double quote. For example: http_user_agent = "\"request_User-Agent\""
  2. A backtick quote. For example: http_user_agent = `"request_User-Agent"`

stage.label_drop block

The stage.label_drop inner block configures a processing stage that drops labels from incoming log entries.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
valueslist(string)Configures a label_drop processing stage.{}no
river
stage.label_drop {
    values = [ "kubernetes_node_name", "kubernetes_namespace" ]
}

stage.label_keep block

The stage.label_keep inner block configures a processing stage that filters the label set of an incoming log entry down to a subset.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
valueslist(string)Configures a label_keep processing stage.{}no
river
stage.label_keep {
    values = [ "kubernetes_pod_name", "kubernetes_pod_container_name" ]
}

stage.labels block

The stage.labels inner block configures a labels processing stage that can read data from the extracted values map and set new labels on incoming log entries.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
valuesmap(string)Configures a labels processing stage.{}no

In a labels stage, the map’s keys define the label to set and the values are how to look them up. If the value is empty, it is inferred to be the same as the key.

river
stage.labels {
    values = {
      env  = "",         // Sets up an 'env' label, based on the 'env' extracted value.
      user = "username", // Sets up a 'user' label, based on the 'username' extracted value.
    }
}

stage.structured_metadata block

The stage.structured_metadata inner block configures a stage that can read data from the extracted values map and add them to log entries as structured metadata.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
valuesmap(string)Specifies the list of labels to add from extracted values map to log entry.{}no

In a structured_metadata stage, the map’s keys define the label to set and the values are how to look them up. If the value is empty, it is inferred to be the same as the key.

river
stage.structured_metadata {
    values = {
      env  = "",         // Sets up an 'env' property to structured metadata, based on the 'env' extracted value.
      user = "username", // Sets up a 'user' property to structured metadata, based on the 'username' extracted value.
    }
}

stage.limit block

The stage.limit inner block configures a rate-limiting stage that throttles logs based on several options.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
ratenumberThe maximum rate of lines per second that the stage forwards.yes
burstnumberThe maximum number of burst lines that the stage forwards.yes
by_label_namestringThe label to use when rate-limiting on a label name.""no
dropboolWhether to discard or backpressure lines that exceed the rate limit.falseno
max_distinct_labelsnumberThe number of unique values to keep track of when rate-limiting by_label_name.10000no

The rate limiting is implemented as a “token bucket” of size burst, initially full and refilled at rate tokens per second. Each received log entry consumes one token from the bucket. When drop is set to true, incoming entries that exceed the rate-limit are dropped, otherwise they are queued until more tokens are available.

river
stage.limit {
    rate  = 5
    burst = 10
}

If by_label_name is set, then drop must be set to true. This enables the stage to rate-limit not by the number of lines but by the number of labels.

The following example rate-limits entries from each unique namespace value independently. Any entries without the namespace label are not rate-limited. The stage keeps track of up to max_distinct_labels unique values, defaulting at 10000.

river
stage.limit {
    rate  = 10
    burst = 10
    drop  = true

    by_label_name = "namespace"
}

stage.logfmt block

The stage.logfmt inner block configures a processing stage that reads incoming log lines as logfmt and extracts values from them.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
mappingmap(string)Key-value pairs of logmft fields to extract.yes
sourcestringSource of the data to parse as logfmt.""no

The source field defines the source of data to parse as logfmt. When source is missing or empty, the stage parses the log line itself, but it can also be used to parse a previously extracted value.

This stage uses the go-logfmt unmarshaler, so that numeric or boolean types are unmarshalled into their correct form. The stage does not perform any other type conversions. If the extracted value is a complex type, it is treated as a string.

Let’s see how this works on the following log line and stages.

time=2012-11-01T22:08:41+00:00 app=loki level=WARN duration=125 message="this is a log line" extra="user=foo"

stage.logfmt {
    mapping = { "extra" = "" }
}

stage.logfmt {
    mapping = { "username" = "user" }
    source  = "extra"
}

The first stage parses the log line itself and inserts the extra key in the set of extracted data, with the value of user=foo.

The second stage parses the contents of extra and appends the username: foo key-value pair to the set of extracted data.

stage.match block

The stage.match inner block configures a filtering stage that can conditionally either apply a nested set of processing stages or drop an entry when a log entry matches a configurable LogQL stream selector and filter expressions.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
selectorstringThe LogQL stream selector and filter expressions to use.yes
pipeline_namestringA custom name to use for the nested pipeline.""no
actionstringThe action to take when the selector matches the log line. Supported values are "keep" and "drop""keep"no
drop_counter_reasonstringA custom reason to report for dropped lines."match_stage"no

The stage.match block supports a number of stage.* inner blocks, like the top-level block. These are used to construct the nested set of stages to run if the selector matches the labels and content of the log entries. It supports all the same stage.NAME blocks as the in the top level of the loki.process component.

If the specified action is "drop", the metric loki_process_dropped_lines_total is incremented with every line dropped. By default, the reason label is "match_stage", but a custom reason can be provided by using the drop_counter_reason argument.

Let’s see this in action, with the following log lines and stages

{ "time":"2023-01-18T17:08:41+00:00", "app":"foo", "component": ["parser","type"], "level" : "WARN", "message" : "app1 log line" }
{ "time":"2023-01-18T17:08:42+00:00", "app":"bar", "component": ["parser","type"], "level" : "ERROR", "message" : "foo noisy error" }

stage.json {
    expressions = { "appname" = "app" }
}

stage.labels {
    values = { "applbl" = "appname" }
}

stage.match {
    selector = "{applbl=\"foo\"}"

    stage.json {
        expressions = { "msg" = "message" }
    }
}

stage.match {
    selector = "{applbl=\"qux\"}"
    stage.json {
        expressions = { "msg" = "msg" }
    }
}

stage.match {
    selector = "{applbl=\"bar\"} |~ \".*noisy error.*\""
    action   = "drop"

    drop_counter_reason = "discard_noisy_errors"
}

stage.output {
    source = "msg"
}

The first two stages parse the log lines as JSON, decode the app value into the shared extracted map as appname, and use its value as the applbl label.

The third stage uses the LogQL selector to only execute the nested stages on lines where the applbl="foo". So, for the first line, the nested JSON stage adds msg="app1 log line" into the extracted map.

The fourth stage uses the LogQL selector to only execute on lines where applbl="qux"; that means it won’t match any of the input, and the nested JSON stage does not run.

The fifth stage drops entries from lines where applbl is set to ‘bar’ and the line contents matches the regex .*noisy error.*. It also increments the loki_process_dropped_lines_total metric with a label drop_counter_reason="discard_noisy_errors".

The final output stage changes the contents of the log line to be the value of msg from the extracted map. In this case, the first log entry’s content is changed to app1 log line.

stage.metrics block

The stage.metrics inner block configures stage that allows to define and update metrics based on values from the shared extracted map. The created metrics are available at the Agent’s root /metrics endpoint.

The stage.metrics block does not support any arguments and is only configured via a number of nested inner metric.* blocks, one for each metric that should be generated.

The following blocks are supported inside the definition of stage.metrics:

HierarchyBlockDescriptionRequired
metric.countermetric.counterDefines a counter metric.no
metric.gaugemetric.gaugeDefines a gauge metric.no
metric.histogrammetric.histogramDefines a histogram metric.no

metric.counter block

Defines a metric whose value only goes up.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
namestringThe metric name.yes
actionstringThe action to take. Valid actions are set, inc, dec, add, or sub.yes
descriptionstringThe metric’s description and help text.""no
sourcestringKey from the extracted data map to use for the metric. Defaults to the metric name.""no
prefixstringThe prefix to the metric name."loki_process_custom_"no
max_idle_durationdurationMaximum amount of time to wait until the metric is marked as ‘stale’ and removed."5m"no
valuestringIf set, the metric only changes if source exactly matches the value.""no
match_allboolIf set to true, all log lines are counted, without attemptng to match the source to the extracted map.falseno
count_entry_bytesboolIf set to true, counts all log lines bytes.falseno

A counter cannot set both match_all to true and a value. A counter cannot set count_entry_bytes without also setting match_all=true or action=add. The valid action values are inc and add. The inc action increases the metric value by 1 for each log line that passed the filter. The add action converts the extracted value to a positive float and adds it to the metric.

metric.gauge block

Defines a gauge metric whose value can go up or down.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
namestringThe metric name.yes
actionstringThe action to take. Valid actions are inc and add.yes
descriptionstringThe metric’s description and help text.""no
sourcestringKey from the extracted data map to use for the metric. Defaults to the metric name.""no
prefixstringThe prefix to the metric name."loki_process_custom_"no
max_idle_durationdurationMaximum amount of time to wait until the metric is marked as ‘stale’ and removed."5m"no
valuestringIf set, the metric only changes if source exactly matches the value.""no

The valid action values are inc, dec, set, add, or sub. inc and dec increment and decrement the metric’s value by 1 respectively. If set, add, or sub` is chosen, the extracted value must be convertible to a positive float and is set, added to, or subtracted from the metric’s value.

metric.histogram block

Defines a histogram metric whose values are recorded in predefined buckets.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
namestringThe metric name.yes
bucketslist(float)The action to take. Valid actions are set, inc, dec, add, or sub.yes
descriptionstringThe metric’s description and help text.""no
sourcestringKey from the extracted data map to use for the metric. Defaults to the metric name.""no
prefixstringThe prefix to the metric name."loki_process_custom_"no
max_idle_durationdurationMaximum amount of time to wait until the metric is marked as ‘stale’ and removed."5m"no
valuestringIf set, the metric only changes if source exactly matches the value.""no

metrics behavior

If value is not present, all incoming log entries match.

Label values on created metrics can be dynamic, which can cause exported metrics to explode in cardinality or go stale, for example, when a stream stops receiving new logs. To prevent unbounded growth of the /metrics endpoint, any metrics which have not been updated within max_idle_duration are removed. The max_idle_duration must be greater or equal to "1s", and it defaults to "5m".

The metric values extracted from the log data are internally converted to floats. The supported values are the following:

  • integer
  • floating point number
  • string - Two types of string format are supported:
    • Strings that represent floating point numbers, for example, “0.804” is converted to 0.804.
    • Duration format strings. Valid time units are “ns”, “us”, “ms”, “s”, “m”, “h”. A value in this format is converted to a floating point number of seconds, for example, “0.5ms” is converted to 0.0005.
  • boolean:
    • true is converted to 1.
    • false is converted to 0.

The following pipeline creates a counter which increments every time any log line is received by using the match_all parameter. The pipeline creates a second counter which adds the byte size of these log lines by using the count_entry_bytes parameter.

These two metrics disappear after 24 hours if no new entries are received, to avoid building up metrics which no longer serve any use. These two metrics are a good starting point to track the volume of log streams in both the number of entries and their byte size, to identify sources of high-volume or high-cardinality data.

river
stage.metrics {
    metric.counter {
        name        = "log_lines_total"
        description = "total number of log lines"
        prefix      = "my_custom_tracking_"

        match_all         = true
        action            = "inc"
        max_idle_duration = "24h"
    }
}
stage.metrics {
    metric.counter {
        name        = "log_bytes_total"
        description = "total bytes of log lines"
        prefix      = "my_custom_tracking_"

        match_all         = true
        count_entry_bytes = true
        action            = "add"
        max_idle_duration = "24h"
    }
}

Here, the first stage uses a regex to extract text in the format order_status=<string> in the log line. The second stage, defines a counter which increments the successful_orders_total and failed_orders_total based on the previously extracted values.

river
stage.regex {
    expression = "^.* order_status=(?P<order_status>.*?) .*$"
}
stage.metrics {
    metric.counter {
        name        = "successful_orders_total"
        description = "successful orders"
        source      = "order_status"
        value       = "success"
        action      = "inc"
    }
}
stage.metrics {
    metric.counter {
        name        = "failed_orders_total"
        description = "failed orders"
        source      = "order_status"
        value       = "fail"
        action      = "inc"
    }
}

In this example, the first stage extracts text in the format of retries=<value>, from the log line. The second stage creates a gauge whose current metric value is increased by the number extracted from the retries field.

river
stage.regex {
    expression = "^.* retries=(?P<retries>\\d+) .*$"
}
stage.metrics {
    metric.gauge {
        name        = "retries_total"
        description = "total_retries"
        source      = "retries"
        action      = "add"
    }
}

The following example shows a histogram that reads response_time from the extracted map and places it into a bucket, both increasing the count of the bucket and the sum for that particular bucket:

river
stage.metrics {
    metric.histogram {
        name        = "http_response_time_seconds"
        description = "recorded response times"
        source      = "response_time"
        buckets     = [0.001,0.0025,0.005,0.010,0.025,0.050]
    }
}

stage.multiline block

The stage.multiline inner block merges multiple lines into a single block before passing it on to the next stage in the pipeline.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
firstlinestringName from extracted data to use for the log entry.yes
max_wait_timedurationThe maximum time to wait for a multiline block."3s"no
max_linesnumberThe maximum number of lines a block can have.128no

A new block is identified by the RE2 regular expression passed in firstline.

Any line that does not match the expression is considered to be part of the block of the previous match. If no new logs arrive with max_wait_time, the block is sent on. The max_lines field defines the maximum number of lines a block can have. If this is exceeded, a new block is started.

Let’s see how this works in practice with an example stage and a stream of log entries from a Flask web service.

stage.multiline {
    firstline     = "^\\[\\d{4}-\\d{2}-\\d{2} \\d{1,2}:\\d{2}:\\d{2}\\]"
    max_wait_time = "10s"
}

[2023-01-18 17:41:21] "GET /hello HTTP/1.1" 200 -
[2023-01-18 17:41:25] ERROR in app: Exception on /error [GET]
Traceback (most recent call last):
  File "/home/pallets/.pyenv/versions/3.8.5/lib/python3.8/site-packages/flask/app.py", line 2447, in wsgi_app
    response = self.full_dispatch_request()
  File "/home/pallets/.pyenv/versions/3.8.5/lib/python3.8/site-packages/flask/app.py", line 1952, in full_dispatch_request
    rv = self.handle_user_exception(e)
  File "/home/pallets/.pyenv/versions/3.8.5/lib/python3.8/site-packages/flask/app.py", line 1821, in handle_user_exception
    reraise(exc_type, exc_value, tb)
  File "/home/pallets/.pyenv/versions/3.8.5/lib/python3.8/site-packages/flask/_compat.py", line 39, in reraise
    raise value
  File "/home/pallets/.pyenv/versions/3.8.5/lib/python3.8/site-packages/flask/app.py", line 1950, in full_dispatch_request
    rv = self.dispatch_request()
  File "/home/pallets/.pyenv/versions/3.8.5/lib/python3.8/site-packages/flask/app.py", line 1936, in dispatch_request
    return self.view_functions[rule.endpoint](**req.view_args)
  File "/home/pallets/src/deployment_tools/hello.py", line 10, in error
    raise Exception("Sorry, this route always breaks")
Exception: Sorry, this route always breaks
[2023-01-18 17:42:24] "GET /error HTTP/1.1" 500 -
[2023-01-18 17:42:29] "GET /hello HTTP/1.1" 200 -

All ‘blocks’ that form log entries of separate web requests start with a timestamp in square brackets. The stage detects this with the regular expression in firstline to collapse all lines of the traceback into a single block and thus a single Loki log entry.

stage.output block

The stage.output inner block configures a processing stage that reads from the extracted map and changes the content of the log entry that is forwarded to the next component.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
sourcestringName from extracted data to use for the log entry.yes

Let’s see how this works for the following log line and three-stage pipeline:

{"user": "John Doe", "message": "hello, world!"}

stage.json {
    expressions = { "user" = "user", "message" = "message" }
}

stage.labels {
    values = { "user" = "user" }
}

stage.output {
    source = "message"
}

The first stage extracts the following key-value pairs into the shared map:

user: John Doe
message: hello, world!

Then, the second stage adds user="John Doe" to the label set of the log entry, and the final output stage changes the log line from the original JSON to hello, world!.

stage.pack block

The stage.pack inner block configures a transforming stage that replaces the log entry with a JSON object that embeds extracted values and labels with it.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
labelslist(string)The values from the extracted data and labels to pack with the log entry.yes
ingest_timestampboolWhether to replace the log entry timestamp with the time the pack stage runs.`trueno

This stage lets you embed extracted values and labels together with the log line, by packing them into a JSON object. The original message is stored under the _entry key, and all other keys retain their values. This is useful in cases where you do want to keep a certain label or metadata, but you don’t want it to be indexed as a label due to high cardinality.

The querying capabilities of Loki make it easy to still access this data so it can be filtered and aggregated at query time.

For example, consider the following log entry:

log_line: "something went wrong"
labels:   { "level" = "error", "env" = "dev", "user_id" = "f8fas0r" }

and this processing stage:

river
stage.pack {
    labels = ["env", "user_id"]
}

The stage transforms the log entry into the following JSON object, where the two embedded labels are removed from the original log entry:

json
{
  "_entry": "something went wrong",
  "env": "dev",
  "user_id": "f8fas0r",
}

At query time, Loki’s unpack parser can be used to access these embedded labels and replace the log line with the original one stored in the _entry field automatically.

When combining several log streams to use with the pack stage, you can set ingest_timestamp to true to avoid interlaced timestamps and out-of-order ingestion issues.

stage.regex block

The stage.regex inner block configures a processing stage that parses log lines using regular expressions and uses named capture groups for adding data into the shared extracted map of values.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
expressionstringA valid RE2 regular expression. Each capture group must be named.yes
sourcestringName from extracted data to parse. If empty, uses the log message.""no

The expression field needs to be a RE2 regex string. Every matched capture group is added to the extracted map, so it must be named like: (?P<name>re). The name of the capture group is then used as the key in the extracted map for the matched value.

Because of how River strings work, any backslashes in expression must be escaped with a double backslash; for example "\\w" or "\\S+".

If the source is empty or missing, then the stage parses the log line itself. If it’s set, the stage parses a previously extracted value with the same name.

Given the following log line and regex stage, the extracted values are shown below:

2019-01-01T01:00:00.000000001Z stderr P i'm a log message!

stage.regex {
    expression = "^(?s)(?P<time>\\S+?) (?P<stream>stdout|stderr) (?P<flags>\\S+?) (?P<content>.*)$"
}

time: 2019-01-01T01:00:00.000000001Z,
stream: stderr,
flags: P,
content: i'm a log message

On the other hand, if the source value is set, then the regex is applied to the value stored in the shared map under that name.

Let’s see what happens when the following log line is put through this two-stage pipeline:

{"timestamp":"2022-01-01T01:00:00.000000001Z"}

stage.json {
    expressions = { time = "timestamp" }
}
stage.regex {
    expression = "^(?P<year>\\d+)"
    source     = "time"
}

The first stage adds the following key-value pair into the extracted map:

time: 2022-01-01T01:00:00.000000001Z

Then, the regex stage parses the value for time from the shared values and appends the subsequent key-value pair back into the extracted values map:

year: 2022

stage.replace block

The stage.replace inner block configures a stage that parses a log line using a regular expression and replaces the log line contents. Named capture groups in the regex also support adding data into the shared extracted map.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
expressionstringName from extracted data to use for the log entry.yes
sourcestringSource of the data to parse. If empty, it uses the log message.no
replacestringValue replaced by the capture group.no

The source field defines the source of data to parse using expression. When source is missing or empty, the stage parses the log line itself, but it can also be used to parse a previously extracted value. The replaced value is assigned back to the source key.

The expression must be a valid RE2 regex. Every named capture group (?P<name>re) is set into the extracted map with its name.

Because of how River treats backslashes in double-quoted strings, note that all backslashes in a regex expression must be escaped like "\\w*".

Let’s see how this works with the following log line and stage. Since source is omitted, the replacement occurs on the log line itself.

2023-01-01T01:00:00.000000001Z stderr P i'm a log message who has sensitive information with password xyz!

stage.replace {
    expression = "password (\\S+)"
    replace    = "*****"
}

The log line is transformed to

2023-01-01T01:00:00.000000001Z stderr P i'm a log message who has sensitive information with password *****!

If replace is empty, then the captured value is omitted instead.

In the following example, source is defined.

{"time":"2023-01-01T01:00:00.000000001Z", "level": "info", "msg":"11.11.11.11 - \"POST /loki/api/push/ HTTP/1.1\" 200 932 \"-\" \"Mozilla/5.0\"}

stage.json {
    expressions = { "level" = "", "msg" = "" }
}

stage.replace {
    expression = "\\S+ - \"POST (\\S+) .*"
    source     = "msg"
    replace    = "redacted_url"
}

The JSON stage adds the following key-value pairs into the extracted map:

time: 2023-01-01T01:00:00.000000001Z
level: info
msg: "11.11.11.11 - "POST /loki/api/push/ HTTP/1.1" 200 932 "-" "Mozilla/5.0"

The replace stage acts on the msg value. The capture group matches against /loki/api/push and is replaced by redacted_url.

The msg value is finally transformed into:

msg: "11.11.11.11 - "POST redacted_url HTTP/1.1" 200 932 "-" "Mozilla/5.0"

The replace field can use a set of templating functions, by utilizing Go’s text/template package.

Let’s see how this works with named capture groups with a sample log line and stage.

11.11.11.11 - agent [01/Jan/2023:00:00:01 +0200]

stage.replace {
    expression = "^(?P<ip>\\S+) (?P<identd>\\S+) (?P<user>\\S+) \\[(?P<timestamp>[\\w:/]+\\s[+\\-]\\d{4})\\]"
    replace    = "{{ .Value | ToUpper }}"
}

Since source is empty, the regex parses the log line itself and extracts the named capture groups to the shared map of values. The replace field acts on these extracted values and converts them to uppercase:

ip: 11.11.11.11
identd: -
user: FRANK
timestamp: 01/JAN/2023:00:00:01 +0200

and the log line becomes:

11.11.11.11 - FRANK [01/JAN/2023:00:00:01 +0200]

The following list contains available functions with examples of more complex replace fields.

ToLower, ToUpper, Replace, Trim, TrimLeftTrimRight, TrimPrefix, TrimSuffix, TrimSpace, Hash, Sha2Hash, regexReplaceAll, regexReplaceAllLiteral

"{{ if eq .Value \"200\" }}{{ Replace .Value \"200\" \"HttpStatusOk\" -1 }}{{ else }}{{ .Value | ToUpper }}{{ end }}"
"*IP4*{{ .Value | Hash "salt" }}*"

stage.static_labels block

The stage.static_labels inner block configures a static_labels processing stage that adds a static set of labels to incoming log entries.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
valuesmap(string)Configures a static_labels processing stage.{}no
river
stage.static_labels {
    values = {
      foo = "fooval",
      bar = "barval",
    }
}

stage.template block

The stage.template inner block configures a transforming stage that allows users to manipulate the values in the extracted map by using Go’s text/template package syntax. This stage is primarily useful for manipulating and standardizing data from previous stages before setting them as labels in a subsequent stage. Example use cases are replacing spaces with underscores, converting uppercase strings to lowercase, or hashing a value.

The template stage can also create new keys in the extracted map.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
sourcestringName from extracted data to parse. If the key doesn’t exist, a new entry is created.yes
templatestringGo template string to use.yes

The template string can be any valid template that can be used by Go’s text/template. It supports all functions from the sprig package, as well as the following list of custom functions:

ToLower, ToUpper, Replace, Trim, TrimLeftTrimRight, TrimPrefix, TrimSuffix, TrimSpace, Hash, Sha2Hash, regexReplaceAll, regexReplaceAllLiteral

More details on each of these functions can be found in the supported functions section below.

Assuming no data is present on the extracted map, the following stage simply adds the new_key: "hello_world"key-value pair to the shared map.

river
stage.template {
    source   = "new_key"
    template = "hello_world"
}

If the source value exists in the extract fields, its value can be referred to as .Value in the template. The next stage takes the current value of app from the extracted map, converts it to lowercase, and adds a suffix to its value:

river
stage.template {
    source   = "app"
    template = "{{ ToLower .Value }}_some_suffix"
}

Any previously extracted keys are available for template to expand and use. The next stage takes the current values for level, app and module and creates a new key named output_message:

river
stage.template {
    source   = "output_msg"
    template = "{{ .level }} for app {{ ToUpper .app }} in module {{.module}}"
}

A special key named Entry can be used to reference the current line; this can be useful when you need to append/prepend something to the log line, like this snippet:

river
stage.template {
    source   = "message"
    template = "{{.app }}: {{ .Entry }}"
}
stage.output {
    source = "message"
}

Supported functions

In addition to supporting all functions from the sprig package, the template stage supports the following custom functions.

ToLower and ToUpper

ToLower and ToUpper convert the entire string to lowercase and uppercase, respectively.

Examples:

river
stage.template {
    source   = "out"
    template = "{{ ToLower .app }}"
}
stage.template {
    source   = "out"
    template = "{{ .app | ToUpper }}"
}
Replace

The Replace function syntax is defined as {{ Replace <string> <old> <new> <n> }}.

The function returns a copy of the input string, with instances of the <old> argument being replaced by <new>. The function replaces up to <n> non-overlapping instances of the second argument. If <n> is less than zero, there is no limit on the number of replacement. Finally, if <old> is empty, it matches before and after every UTF-8 character in the string.

This example replaces the first two instances of the loki word by Loki:

river
stage.template {
    source   = "output"
    template = "{{ Replace .Value "loki" "Loki" 2 }}"
}
Trim, TrimLeft, TrimRight, TrimSpace, TrimPrefix, TrimSuffix
  • Trim returns a slice of the string s with all leading and trailing Unicode code points contained in cutset removed.
  • TrimLeft and TrimRight are the same as Trim except that they trim only leading and trailing characters, respectively.
  • TrimSpace returns a slice of the string s, with all leading and trailing white space removed, as defined by Unicode.
  • TrimPrefix and TrimSuffix trim the supplied prefix or suffix, respectively. Examples:
river
stage.template {
    source   = "output"
    template = "{{ Trim .Value ",. " }}"
}
stage.template {
    source   = "output"
    template = "{{ TrimSpace .Value }}"
}
stage.template {
    source   = "output"
    template = "{{ TrimPrefix .Value "--" }}"
}
Regex

regexReplaceAll returns a copy of the input string, replacing matches of the Regexp with the replacement string. Inside the replacement string, $ characters are interpreted as in Expand functions, so for instance, $1 represents the first captured submatch.

regexReplaceAllLiteral returns a copy of the input string, replacing matches of the Regexp with the replacement string. The replacement string is substituted directly, without using Expand.

river
stage.template {
    source   = "output"
    template = "{{ regexReplaceAll "(a*)bc" .Value "${1}a" }}"
}
stage.template {
    source   = "output"
    template = "{{ regexReplaceAllLiteral "(ts=)" .Value "timestamp=" }}"
}
Hash and Sha2Hash

Hash returns a Sha3_256 hash of the string, represented as a hexadecimal number of 64 digits. You can use it to obfuscate sensitive data and PII in the logs. It requires a (fixed) salt value, to add complexity to low input domains (e.g., all possible social security numbers). Sha2Hash returns a Sha2_256 of the string which is faster and less CPU-intensive than Hash, however it is less secure.

Examples:

river
stage.template {
    source   = "output"
    template = "{{ Hash .Value "salt" }}"
}
stage.template {
    source   = "output"
    template = "{{ Sha2Hash .Value "salt" }}"
}

We recommend using Hash as it has a stronger hashing algorithm.

stage.tenant block

The stage.tenant inner block sets the tenant ID for the log entry by obtaining it from a field in the extracted data map, a label, or a provided value.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
labelstringThe label to set as tenant ID.""no
sourcestringThe name from the extracted value to use as tenant ID.""no
valuestringThe value to set as the tenant ID.""no

The block expects only one of label, source or value to be provided.

The following stage assigns the fixed value team-a as the tenant ID:

river
stage.tenant {
    value = "team-a"
}

This stage extracts the tenant ID from the customer_id field after parsing the log entry as JSON in the shared extracted map:

river
stage.json {
    expressions = { "customer_id" = "" }
}
stage.tenant {
    source = "customer_id"
}

The final example extracts the tenant ID from a label set by a previous stage:

river
stage.labels {
    "namespace" = "k8s_namespace"
}
stage.tenant {
    label = "namespace"
}

stage.timestamp block

The stage.timestamp inner block configures a processing stage that sets the timestamp of log entries before they’re forwarded to the next component. When no timestamp stage is set, the log entry timestamp defaults to the time when the log entry was scraped.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
sourcestringName from extracted values map to use for the timestamp.yes
formatstringDetermines how to parse the source string.yes
fallback_formatslist(string)Fallback formats to try if the format field fails.[]no
locationstringIANA Timezone Database location to use when parsing.""no
action_on_failurestringWhat to do when the timestamp can’t be extracted or parsed."fudge"no

The source field defines which value from the shared map of extracted values the stage should attempt to parse as a timestamp.

The format field defines how that source should be parsed.

First off, the format can be set to one of the following shorthand values for commonly-used forms:

ANSIC: Mon Jan _2 15:04:05 2006
UnixDate: Mon Jan _2 15:04:05 MST 2006
RubyDate: Mon Jan 02 15:04:05 -0700 2006
RFC822: 02 Jan 06 15:04 MST
RFC822Z: 02 Jan 06 15:04 -0700
RFC850: Monday, 02-Jan-06 15:04:05 MST
RFC1123: Mon, 02 Jan 2006 15:04:05 MST
RFC1123Z: Mon, 02 Jan 2006 15:04:05 -0700
RFC3339: 2006-01-02T15:04:05-07:00
RFC3339Nano: 2006-01-02T15:04:05.999999999-07:00

Additionally, support for common Unix timestamps is supported with the following format values:

Unix: 1562708916 or with fractions 1562708916.000000123
UnixMs: 1562708916414
UnixUs: 1562708916414123
UnixNs: 1562708916000000123

Otherwise, the field accepts a custom format string that defines how an arbitrary reference point in history should be interpreted by the stage. The arbitrary reference point is Mon Jan 2 15:04:05 -0700 MST 2006.

The string value of the field is passed directly to the layout parameter in Go’s time.Parse function.

If the custom format has no year component, the stage uses the current year, according to the system’s clock.

The following table shows the supported reference values to use when defining a custom format.

Timestamp ComponentFormat value
Year06, 2006
Month1, 01, Jan, January
Day2, 02, _2 (two digits right justified)
Day of the weekMon, Monday
Hour3 (12-hour), 03 (12-hour zero prefixed), 15 (24-hour)
Minute4, 04
Second5, 05
Fraction of second.000 (ms zero prefixed), .000000 (μs), .000000000 (ns), .999 (ms without trailing zeroes), .999999 (μs), .999999999 (ns)
12-hour periodpm, PM
Timezone nameMST
Timezone offset-0700, -070000 (with seconds), -07, 07:00, -07:00:00 (with seconds)
Timezone ISO-8601Z0700 (Z for UTC or time offset), Z070000, Z07, Z07:00, Z07:00:00

The fallback_formats field defines one or more format fields to try and parse the timestamp with, if parsing with format fails.

The location field must be a valid IANA Timezone Database location and determines in which timezone the timestamp value is interpreted to be in.

The action_on_failure field defines what should happen when the source field doesn’t exist in the shared extracted map, or if the timestamp parsing fails.

The supported actions are:

  • fudge (default): Change the timestamp to the last known timestamp, summing up 1 nanosecond (to guarantee log entries ordering).
  • skip: Do not change the timestamp and keep the time when the log entry was scraped.

The following stage fetches the time value from the shared values map, parses it as a RFC3339 format, and sets it as the log entry’s timestamp.

river
stage.timestamp {
    source = "time"
    format = "RFC3339"
}

stage.geoip block

The stage.geoip inner block configures a processing stage that reads an IP address and populates the shared map with geoip fields. Maxmind’s GeoIP2 database is used for the lookup.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
dbstringPath to the Maxmind DB file.yes
sourcestringIP from extracted data to parse.yes
db_typestringMaxmind DB type. Allowed values are “city”, “asn”.no
custom_lookupsmap(string)Key-value pairs of JMESPath expressions.no

GeoIP with City database example:

{"log":"log message","client_ip":"34.120.177.193"}

loki.process "example" {
	stage.json {
		expressions = {ip = "client_ip"}
	}

	stage.geoip {
		source  = "ip"
		db      = "/path/to/db/GeoLite2-City.mmdb"
		db_type = "city"
	}

	stage.labels {
		values = {
			geoip_city_name          = "",
			geoip_country_name       = "",
			geoip_continent_name     = "",
			geoip_continent_code     = "",
			geoip_location_latitude  = "",
			geoip_location_longitude = "",
			geoip_postal_code        = "",
			geoip_timezone           = "",
			geoip_subdivision_name   = "",
			geoip_subdivision_code   = "",
		}
	}
}

The json stage extracts the IP address from the client_ip key in the log line. Then the extracted ip value is given as source to geoip stage. The geoip stage performs a lookup on the IP and populates the following fields in the shared map which are added as labels using the labels stage.

The extracted data from the IP used in this example:

  • geoip_city_name: Kansas City
  • geoip_country_name: United States
  • geoip_continet_name: North America
  • geoip_continent_code: NA
  • geoip_location_latitude: 39.1027
  • geoip_location_longitude: -94.5778
  • geoip_postal_code: 64184
  • geoip_timezone: America/Chicago
  • geoip_subdivision_name: Missouri
  • geoip_subdivision_code: MO

GeoIP with ASN (Autonomous System Number) database example

loki.process "example" {
	stage.json {
		expressions = {ip = "client_ip"}
	}

	stage.geoip {
		source  = "ip"
		db      = "/path/to/db/GeoIP2-ASN.mmdb"
		db_type = "asn"
	}

	stage.labels {
		values = {
			geoip_autonomous_system_number       = "",
			geoip_autonomous_system_organization = "",
		}
	}
}

The json stage extracts the IP address from the client_ip key in the log line. Then the extracted ip value is given as source to geoip stage. The geoip stage performs a lookup on the IP and populates the shared map.

The extracted data from the IP used in this example:

  • geoip_autonomous_system_number: 396982
  • geoip_autonomous_system_organization: GOOGLE-CLOUD-PLATFORM

GeoIP with custom fields example

If the MMDB file used is enriched with custom data, for example, private IP addresses as explained in the Maxmind blog post, then it can be extracted from the record using the custom_lookups attribute.

loki.process "example" {
	stage.json {
		expressions = {ip = "client_ip"}
	}

	stage.geoip {
		source         = "ip"
		db             = "/path/to/db/GeoIP2-Enriched.mmdb"
		db_type        = "city"
		custom_lookups = {
			"department"  = "MyCompany.DeptName",
			"parent_vnet" = "MyCompany.ParentVNet",
			"subnet"      = "MyCompany.Subnet",
		}
	}

	stage.labels {
		values = {
			department  = "",
			parent_vnet = "",
			subnet      = "",
		}
	}
}

The json stage extracts the IP address from the client_ip key in the log line. Then the extracted ip value is given as source to geoip stage. The geoip stage performs a lookup on the IP and populates the shared map with the data from the city database results in addition to the custom lookups. Lastly, the custom lookup fields from the shared map are added as labels.

Exported fields

The following fields are exported and can be referenced by other components:

NameTypeDescription
receiverLogsReceiverA value that other components can use to send log entries to.

Component health

loki.process is only reported as unhealthy if given an invalid configuration.

Debug information

loki.process does not expose any component-specific debug information.

Debug metrics

  • loki_process_dropped_lines_total (counter): Number of lines dropped as part of a processing stage.
  • loki_process_dropped_lines_by_label_total (counter): Number of lines dropped when by_label_name is non-empty in stage.limit.

Example

This example creates a loki.process component that extracts the environment value from a JSON log line and sets it as a label named ’env’.

river
loki.process "local" {
  forward_to = [loki.write.onprem.receiver]

  stage.json {
      expressions = { "extracted_env" = "environment" }
  }

  stage.labels {
      values = { "env" = "extracted_env" }
  }
}