oliver006/redis_exporter

Prometheus Redis Metrics Exporter

Prometheus exporter for Redis metrics.\
Supports Redis 2.x, 3.x, 4.x, 5.x, 6.x, and 7.x

Ukraine is currently suffering from Russian aggression, please consider supporting Ukraine with a donation.

Building and running the exporter

Build and run locally

git clone https://github.com/oliver006/redis_exporter.git
cd redis_exporter
go build .
./redis_exporter --version

Pre-build binaries

For pre-built binaries please take a look at the releases.

Basic Prometheus Configuration

Add a block to the scrape_configs of your prometheus.yml config file:

scrape_configs:
  - job_name: redis_exporter
    static_configs:
    - targets: ['<<REDIS-EXPORTER-HOSTNAME>>:9121']

and adjust the host name accordingly.

Kubernetes SD configurations

To have instances in the drop-down as human readable names rather than IPs, it is suggested to use instance relabelling.

For example, if the metrics are being scraped via the pod role, one could add:

          - source_labels: [__meta_kubernetes_pod_name]
            action: replace
            target_label: instance
            regex: (.*redis.*)

as a relabel config to the corresponding scrape config. As per the regex value, only pods with "redis" in their name will be relabelled as such.

Similar approaches can be taken with other role types depending on how scrape targets are retrieved.

Prometheus Configuration to Scrape Multiple Redis Hosts

Run the exporter with the command line flag --redis.addr= so it won't try to access the local instance every time the /metrics endpoint is scraped. Using below config instead of the /metric endpoint the /scrape endpoint will be used by prometheus. As an example the first target will be queried with this web request:
http://exporterhost:9121/scrape?target=first-redis-host:6379

scrape_configs:
  ## config for the multiple Redis targets that the exporter will scrape
  - job_name: 'redis_exporter_targets'
    static_configs:
      - targets:
        - redis://first-redis-host:6379
        - redis://second-redis-host:6379
        - redis://second-redis-host:6380
        - redis://second-redis-host:6381
    metrics_path: /scrape
    relabel_configs:
      - source_labels: [__address__]
        target_label: __param_target
      - source_labels: [__param_target]
        target_label: instance
      - target_label: __address__
        replacement: <<REDIS-EXPORTER-HOSTNAME>>:9121

  ## config for scraping the exporter itself
  - job_name: 'redis_exporter'
    static_configs:
      - targets:
        - <<REDIS-EXPORTER-HOSTNAME>>:9121

The Redis instances are listed under targets, the Redis exporter hostname is configured via the last relabel_config rule.\
If authentication is needed for the Redis instances then you can set the password via the --redis.password command line option of
the exporter (this means you can currently only use one password across the instances you try to scrape this way. Use several
exporters if this is a problem). \
You can also use a json file to supply multiple targets by using file_sd_configs like so:

scrape_configs:
  - job_name: 'redis_exporter_targets'
    file_sd_configs:
      - files:
        - targets-redis-instances.json
    metrics_path: /scrape
    relabel_configs:
      - source_labels: [__address__]
        target_label: __param_target
      - source_labels: [__param_target]
        target_label: instance
      - target_label: __address__
        replacement: <<REDIS-EXPORTER-HOSTNAME>>:9121

  ## config for scraping the exporter itself
  - job_name: 'redis_exporter'
    static_configs:
      - targets:
        - <<REDIS-EXPORTER-HOSTNAME>>:9121

The targets-redis-instances.json should look something like this:

[
  {
    "targets": [ "redis://redis-host-01:6379", "redis://redis-host-02:6379"],
    "labels": { }
  }
]

Prometheus uses file watches and all changes to the json file are applied immediately.

Command line flags

Redis instance addresses can be tcp addresses: redis://localhost:6379, redis.example.com:6379 or e.g. unix sockets: unix:///tmp/redis.sock.\
SSL is supported by using the rediss:// schema, for example: rediss://azure-ssl-enabled-host.redis.cache.windows.net:6380 (note that the port is required when connecting to a non-standard 6379 port, e.g. with Azure Redis instances).\

Command line settings take precedence over any configurations provided by the environment variables.

Authenticating with Redis

If your Redis instance requires authentication then there are several ways how you can supply
a username (new in Redis 6.x with ACLs) and a password.

You can provide the username and password as part of the address, see here for the official documentation of the redis:// scheme.
You can set -redis.password-file=sample-pwd-file.json to specify a password file, it's used whenever the exporter connects to a Redis instance,
no matter if you're using the /scrape endpoint for multiple instances or the normal /metrics endpoint when scraping just one instance.
It only takes effect when redis.password == "". See the contrib/sample-pwd-file.json for a working example, and make sure to always include the redis:// in your password file entries.

An example for a URI including a password is: redis://<<username (optional)>>:<<PASSWORD>>@<<HOSTNAME>>:<<PORT>>

Alternatively, you can provide the username and/or password using the --redis.user and --redis.password directly to the redis_exporter.

If you want to use a dedicated Redis user for the redis_exporter (instead of the default user) then you need enable a list of commands for that user.
You can use the following Redis command to set up the user, just replace <<<USERNAME>>> and <<<PASSWORD>>> with your desired values.

ACL SETUSER <<<USERNAME>>> +client +ping +info +config|get +cluster|info +slowlog +latency +memory +select +get +scan +xinfo +type +pfcount +strlen +llen +scard +zcard +hlen +xlen +eval allkeys on > <<<PASSWORD>>>

Run via Docker

The latest release is automatically published to the Docker registry.

You can run it like this:

docker run -d --name redis_exporter -p 9121:9121 oliver006/redis_exporter

Docker images are also published to the quay.io docker repo so you can pull them from there if for instance you run into rate limiting issues with Docker hub.

docker run -d --name redis_exporter -p 9121:9121 quay.io/oliver006/redis_exporter

The latest docker image contains only the exporter binary.
If e.g. for debugging purposes, you need the exporter running
in an image that has a shell then you can run the alpine image:

docker run -d --name redis_exporter -p 9121:9121 oliver006/redis_exporter:alpine

If you try to access a Redis instance running on the host node, you'll need to add --network host so the
redis_exporter container can access it:

docker run -d --name redis_exporter --network host oliver006/redis_exporter

Run on Kubernetes

Here is an example Kubernetes deployment configuration for how to deploy the redis_exporter as a sidecar to a Redis instance.

Tile38

Tile38 now has native Prometheus support for exporting server metrics and basic stats about number of objects, strings, etc.
You can also use redis_exporter to export Tile38 metrics, especially more advanced metrics by using Lua scripts or the -check-keys flag.\
To enable Tile38 support, run the exporter with --is-tile38=true.

What's exported

Most items from the INFO command are exported,
see Redis documentation for details.\
In addition, for every database there are metrics for total keys, expiring keys and the average TTL for keys in the database.\
You can also export values of keys by using the -check-keys (or related) flag. The exporter will also export the size (or, depending on the data type, the length) of the key.
This can be used to export the number of elements in (sorted) sets, hashes, lists, streams, etc.
If a key is in string format and matches with --check-keys (or related) then its string value will be exported as a label in the key_value_as_string metric.

If you require custom metric collection, you can provide a Redis Lua script using the -script flag. An example can be found in the contrib folder.

The redis_memory_max_bytes metric

The metric redis_memory_max_bytes will show the maximum number of bytes Redis can use.\
It is zero if no memory limit is set for the Redis instance you're scraping (this is the default setting for Redis).\
You can confirm that's the case by checking if the metric redis_config_maxmemory is zero or by connecting to the Redis instance via redis-cli and running the command CONFIG GET MAXMEMORY.

What it looks like

Example Grafana screenshots:

Grafana dashboard is available on grafana.com and/or github.com.

Viewing multiple Redis simultaneously

If running Redis Sentinel, it may be desirable to view the metrics of the various cluster members simultaneously. For this reason the dashboard's drop down is of the multi-value type, allowing for the selection of multiple Redis. Please note that there is a caveat; the single stat panels up top namely uptime, total memory use and clients do not function upon viewing multiple Redis.

Using the mixin

There is a set of sample rules, alerts and dashboards available in redis-mixin

Upgrading from 0.x to 1.x

PR #256 introduced breaking changes which were released as version v1.0.0.

If you only scrape one Redis instance and use command line flags --redis.address
and --redis.password then you're most probably not affected.
Otherwise, please see PR #256 and this README for more information.

Memory Usage Aggregation by Key Groups

When a single Redis instance is used for multiple purposes, it is useful to be able to see how Redis memory is consumed among the different usage scenarios. This is particularly important when a Redis instance with no eviction policy is running low on memory as we want to identify whether certain applications are misbehaving (e.g. not deleting keys that are no longer in use) or the Redis instance needs to be scaled up to handle the increased resource demand. Fortunately, most applications using Redis will employ some sort of naming conventions for keys tied to their specific purpose such as (hierarchical) namespace prefixes which can be exploited by the check-keys, check-single-keys, and count-keys parameters of redis_exporter to surface the memory usage metrics of specific scenarios. Memory usage aggregation by key groups takes this one step further by harnessing the flexibility of Redis LUA scripting support to classify all keys on a Redis instance into groups through a list of user-defined LUA regular expressions so memory usage metrics can be aggregated into readily identifiable groups.

To enable memory usage aggregation by key groups, simply specify a non-empty comma-separated list of LUA regular expressions through the check-key-groups redis_exporter parameter. On each aggregation of memory metrics by key groups, redis_exporter will set up a SCAN cursor through all keys for each Redis database to be processed in batches via a LUA script. Each key batch is then processed by the same LUA script on a key-by-key basis as follows:

1. The MEMORY USAGE command is called to gather memory usage for each key
2. The specified LUA regexes are applied to each key in the specified order, and the group name that a given key belongs to will be derived from concatenating the capture groups of the first regex that matches the key. For example, applying the regex ^(.*)_[^_]+$ to the key key_exp_Nick would yield a group name of key_exp. If none of the specified regexes matches a key, the key will be assigned to the unclassified group

Once a key has been classified, the memory usage and key counter for the corresponding group will be incremented in a local LUA table. This aggregated metrics table will then be returned alongside the next SCAN cursor position to redis_exporter when all keys in a batch have been processed, and redis_exporter can aggregate the data from all batches into a single table of grouped memory usage metrics for the Prometheus metrics scrapper.

Besides making the full flexibility of LUA regex available for classifying keys into groups, the LUA script also has the benefit of reducing network traffic by executing all MEMORY USAGE commands on the Redis server and returning aggregated data to redis_exporter in a far more compact format than key-level data. The use of SCAN cursor over batches of keys processed by a server-side LUA script also helps prevent unbounded latency bubble in Redis's single processing thread, and the batch size can be tailored to specific environments via the check-keys-batch-size parameter.

Scanning the entire key space of a Redis instance may sound a lttle extravagant, but it takes only a single scan to classify all keys into groups, and on a moderately sized system with ~780K keys and a rather complex list of 17 regexes, it takes an average of ~5s to perform a full aggregation of memory usage by key groups. Of course, the actual performance for specific systems will vary widely depending on the total number of keys, the number and complexity of regexes used for classification, and the configured batch size.

To protect Prometheus from being overwhelmed by a large number of time series resulting from misconfigured group classification regular expression (e.g. applying the regular expression ^(.*)$ where each key will be classified into its own distinct group), a limit on the number of distinct key groups per Redis database can be configured via the max-distinct-key-groups parameter. If the max-distinct-key-groups limit is exceeded, only the key groups with the highest memory usage within the limit will be tracked separately, remaining key groups will be reported under a single overflow key group.

Here is a list of additional metrics that will be exposed when memory usage aggregation by key groups is enabled:

Script to collect Redis lists and respective sizes.

If using Redis version < 4.0, most of the helpful metrics which we need to gather based on length or memory is not possible via default redis_exporter.
With the help of LUA scripts, we can gather these metrics.
One of these scripts contrib/collect_lists_length_growing.lua will help to collect the length of redis lists.
With this count, we can take following actions such as Create alerts or dashboards in Grafana or any similar tools with these Prometheus metrics.

Development

The tests require a variety of real Redis instances to not only verify correctness of the exporter but also
compatibility with older versions of Redis and with Redis-like systems like KeyDB or Tile38.\
The contrib/docker-compose-for-tests.yml file has service definitions for
everything that's needed.\
You can bring up the Redis test instances first by running make docker-env-up and then, every time you want to run the tests, you can run make docker-test. This will mount the current directory (with the .go source files) into a docker container and kick off the tests.\
Once you're done testing you can bring down the stack by running make docker-env-down.\
Or you can bring up the stack, run the tests, and then tear down the stack, all in one shot, by running make docker-all.

Note. Tests initialization can lead to unexpected results when using a persistent testing environment. When make docker-env-up is executed once and make docker-test is constantly run or stopped during execution, the number of keys in the database changes, which can lead to unexpected failures of tests. Use make docker-env-down periodacally to clean up as a workaround.

Communal effort

Open an issue or PR if you have more suggestions, questions or ideas about what to add.

Redis Exporter Helm Chart

The exporter, alert rule, and dashboard can be deployed in Kubernetes using the Helm chart. The Helm chart used for deployment is taken from the Prometheus community, which can be found here.

Installing Redis cluster

If your Redis cluster is not up and ready, you can start it using Helm:

$ helm repo add bitnami https://charts.bitnami.com/bitnami
$ helm install my-release  bitnami/redis --set master.extraFlags={"--maxmemory 1gb"}

Note that bitnami charts allow you to start a Redis exporter as a side car for the Redis container. You can enable that by adding “--set metrics.enabled=true”

Installing Redis Exporter

helm repo add Prometheus-community https://prometheus-community.github.io/helm-charts

helm repo update
helm install my-release prometheus-community/prometheus-redis-exporter

Some of the common parameters that must be changed in the values file include: 

redisAddress: "redis://redis-master:6379"
Auth:
  enabled: true
  redisPassword: secretpassword 

All these parameters can be tuned via the values.yaml file here.

Scrape the metrics

There are multiple ways to scrape the metrics as discussed above. In addition to the native way of setting up Prometheus monitoring, a service monitor can be deployed (if a Prometheus operator is being used) to scrap the data from the Redis exporter. With this approach, multiple Redis servers can be scrapped without altering the Prometheus configuration. Every Redis exporter comes with its own service monitor.
In the above-mentioned chart, a service monitor can be deployed by turning it on from the values.yaml file here.

serviceMonitor:
  # When set true then use a ServiceMonitor to configure scraping
  enabled: true
  # Set the namespace the ServiceMonitor should be deployed
  # namespace: monitoring
  # Set how frequently Prometheus should scrape
  # interval: 30s
  # Set path to redis-exporter telemtery-path
  # telemetryPath: /metrics
  # Set labels for the ServiceMonitor, use this to define your scrape label for Prometheus Operator
  # labels:
  # Set timeout for scrape
  # timeout: 10s
  # Set relabel_configs as per https://prometheus.io/docs/prometheus/latest/configuration/configuration/#relabel_config
  # relabelings: []
  # Set of labels to transfer on the Kubernetes Service onto the target.
  # targetLabels: []
  # metricRelabelings: []

Update the annotation section here if not using the Prometheus Operator.

service: 
  annotations:
    prometheus.io/path: /metrics
    prometheus.io/scrape: "true"

This concludes our review of the Redis exporter! If you have any questions, you can reach to us via support@nexclipper.io for further discussions. Stay tuned for more useful exporter reviews and other tips coming soon.

Redis Exporter Alerts

After digging into all the valuable metrics, this section explains in detail how we can get critical alerts.

PromQL is a query language for the Prometheus monitoring system. It is designed for building powerful yet simple queries for graphs, alerts, or derived time series (aka recording rules). PromQL is designed from scratch and has zero common grounds with other query languages used in time series databases, such as SQL in TimescaleDB, InfluxQL, or Flux. More details can be found here.

Prometheus comes with a built-in Alert Manager that is responsible for sending alerts (could be email, Slack, or any other supported channel) when any of the trigger conditions is met. Alerting rules allow users to define alerts based on Prometheus query expressions. They are defined based on the available metrics scraped by the exporter. Click here for a good source for community-defined alerts.

A general alert looks as follows:

- alert:(Alert Name)
expr: (Metric exported from exporter) >/</==/<=/=> (Value)
for: (wait for a certain duration between first encountering a new expression output vector element and counting an alert as firing for this element)
labels: (allows specifying a set of additional labels to be attached to the alert)
annotation: (specifies a set of informational labels that can be used to store longer additional information)

Some of the recommended Redit alerts are:

➡ Alert - Redis is down

 - alert: RedisDown
    expr: redis_up == 0
    for: 0m
    labels:
      severity: critical
    annotations:
      summary: Redis down (instance {{ $labels.instance }})
      description: "Redis instance is down\n  VALUE = {{ $value }}\n  LABELS = {{ $labels }}"

➡ Alert - Redis out of memory

# The exporter must be started with --include-system-metrics flag or REDIS_EXPORTER_INCL_SYSTEM_METRICS=true environment variable.
  - alert: RedisOutOfSystemMemory
    expr: redis_memory_used_bytes / redis_total_system_memory_bytes * 100 > 90
    for: 2m
    labels:
      severity: warning
    annotations:
      summary: Redis out of system memory (instance {{ $labels.instance }})
      description: "Redis is running out of system memory (> 90%)\n  VALUE = {{ $value }}\n  LABELS = {{ $labels }}"

➡ Alert - Too many connections

- alert: RedisTooManyConnections
    expr: redis_connected_clients > 100
    for: 2m
    labels:
      severity: warning
    annotations:
      summary: Redis too many connections (instance {{ $labels.instance }})
      description: "Redis instance has too many connections\n  VALUE = {{ $value }}\n  LABELS = {{ $labels }}"

➡ Alert -  Redis rejecting connections

 - alert: RedisRejectedConnections
    expr: increase(redis_rejected_connections_total[1m]) > 0
    for: 0m
    labels:
      severity: critical
    annotations:
      summary: Redis rejected connections (instance {{ $labels.instance }})
      description: "Some connections to Redis has been rejected\n  VALUE = {{ $value }}\n  LABELS = {{ $labels }}"

➡ Alert - Redis out of system memory

# The exporter must be started with --include-system-metrics flag or REDIS_EXPORTER_INCL_SYSTEM_METRICS=true environment variable.
  - alert: RedisOutOfSystemMemory
    expr: redis_memory_used_bytes / redis_total_system_memory_bytes * 100 > 90
    for: 2m
    labels:
      severity: warning
    annotations:
      summary: Redis out of system memory (instance {{ $labels.instance }})
      description: "Redis is running out of system memory (> 90%)\n  VALUE = {{ $value }}\n  LABELS = {{ $labels }}"

➡ Redis missing master

 - alert: RedisMissingMaster
    expr: (count(redis_instance_info{role="master"}) or vector(0)) < 1
    for: 0m
    labels:
      severity: critical
    annotations:
      summary: Redis missing master (instance {{ $labels.instance }})
      description: "Redis cluster has no node marked as master.\n  VALUE = {{ $value }}\n  LABELS = {{ $labels }}"

➡ Redis disconnected slaves

 - alert: RedisDisconnectedSlaves
    expr: count without (instance, job) (redis_connected_slaves) - sum without (instance, job) (redis_connected_slaves) - 1 > 1
    for: 0m
    labels:
      severity: critical
    annotations:
      summary: Redis disconnected slaves (instance {{ $labels.instance }})
      description: "Redis not replicating for all slaves. Consider reviewing the redis replication status.\n  VALUE = {{ $value }}\n  LABELS = {{ $labels }}"

➡ Redis replication broken

 - alert: RedisReplicationBroken
    expr: delta(redis_connected_slaves[1m]) < 0
    for: 0m
    labels:
      severity: critical
    annotations:
      summary: Redis replication broken (instance {{ $labels.instance }})
      description: "Redis instance lost a slave\n  VALUE = {{ $value }}\n  LABELS = {{ $labels }}"

Redis Exporter Grafana

Graphs are easier to understand and more user-friendly than a row of numbers. For this purpose, users can plot their time series data in visualized format using Grafana.

Grafana is an open-source dashboarding tool used for visualizing metrics with the help of customizable and illustrative charts and graphs. It connects very well with Prometheus and makes monitoring easy and informative. Dashboards in Grafana are made up of panels, with each panel running a PromQL query to fetch metrics from Prometheus.
Grafana supports community-driven graphs for most of the widely used software, which can be directly imported to the Grafana Community.

NexClipper uses the Redis by the downager dashboard, which is widely accepted and has a lot of useful panels.

What is a Panel?

Panels are the most basic component of a dashboard and can display information in various ways, such as gauge, text, bar chart, graph, and so on. They provide information in a very interactive way. Users can view every panel separately and check the value of metrics within a specific time range. 
The values on the panel are queried using PromQL, which is Prometheus Query Language. PromQL is a simple query language used to query metrics within Prometheus. It enables users to query data, aggregate and apply arithmetic functions to the metrics, and then further visualize them on panels.

Here are some examples of panels: