SidekiqAlive offers a solution to add liveness probe for a Sidekiq instance deployed in Kubernetes. This library can be used to check sidekiq health outside kubernetes.
How?
A http server is started and on each requests validates that a liveness key is stored in Redis. If it is there means is working.
A Sidekiq worker is the responsible to storing this key. If Sidekiq stops processing workers this key gets expired by Redis an consequently the http server will return a 500 error.
This worker is responsible to requeue itself for the next liveness probe.
Each instance in kubernetes will be checked based on ENV
variable HOSTNAME
(kubernetes sets this for each replica/pod).
On initialization SidekiqAlive will asign to Sidekiq::Worker a queue with the current host and add this queue to the current instance queues to process.
example:
hostname: foo
Worker queue: sidekiq_alive-foo
instance queues:
- sidekiq_alive-foo
*- your queues
hostname: bar
Worker queue: sidekiq_alive-bar
instance queues:
- sidekiq_alive-bar
*- your queues
Add this line to your application's Gemfile:
gem 'sidekiq_alive'
And then execute:
$ bundle
Or install it yourself as:
$ gem install sidekiq_alive
SidekiqAlive will start when running sidekiq
command.
Run Sidekiq
bundle exec sidekiq
curl localhost:7433
#=> Alive!
how to disable?
You can disabled by setting ENV
variable DISABLE_SIDEKIQ_ALIVE
example:
DISABLE_SIDEKIQ_ALIVE=true bundle exec sidekiq
Set livenessProbe
in your Kubernetes deployment
example with recommended setup:
spec:
containers:
- name: my_app
image: my_app:latest
env:
- name: RAILS_ENV
value: production
command:
- bundle
- exec
- sidekiq
ports:
- containerPort: 7433
livenessProbe:
httpGet:
path: /
port: 7433
initialDelaySeconds: 80 # app specific. Time your sidekiq takes to start processing.
timeoutSeconds: 5 # can be much less
readinessProbe:
httpGet:
path: /
port: 7433
initialDelaySeconds: 80 # app specific
timeoutSeconds: 5 # can be much less
lifecycle:
preStop:
exec:
# SIGTERM triggers a quick exit; gracefully terminate instead
command: ['bundle', 'exec', 'sidekiqctl', 'quiet']
terminationGracePeriodSeconds: 60 # put your longest Job time here plus security time.
Create file:
_kube/sidekiqquiet
#!/bin/bash
# Find Pid
SIDEKIQ_PID=$(ps aux | grep sidekiq | grep busy | awk '{ print $2 }')
# Send TSTP signal
kill -SIGTSTP $SIDEKIQ_PID
Make it executable:
$ chmod +x kube/sidekiq_quiet
Execute it in your deployment preStop:
spec:
containers:
- name: my_app
image: my_app:latest
env:
- name: RAILS_ENV
value: production
command:
- bundle
- exec
- sidekiq
ports:
- containerPort: 7433
livenessProbe:
httpGet:
path: /
port: 7433
initialDelaySeconds: 80 # app specific. Time your sidekiq takes to start processing.
timeoutSeconds: 5 # can be much less
readinessProbe:
httpGet:
path: /
port: 7433
initialDelaySeconds: 80 # app specific
timeoutSeconds: 5 # can be much less
lifecycle:
preStop:
exec:
# SIGTERM triggers a quick exit; gracefully terminate instead
command: ['kube/sidekiq_quiet']
terminationGracePeriodSeconds: 60 # put your longest Job time here plus security time.
It's just up to you how you want to use it.
An example in local would be:
bundle exec sidekiq
# let it initialize ...
curl localhost:7433
#=> Alive!
SidekiqAlive.setup do |config|
# ==> Server host
# Host to bind the server.
# Can also be set with the environment variable SIDEKIQ_ALIVE_HOST.
# default: 0.0.0.0
#
# config.host = 0.0.0.0
# ==> Server port
# Port to bind the server.
# Can also be set with the environment variable SIDEKIQ_ALIVE_PORT.
# default: 7433
#
# config.port = 7433
# ==> Server path
# HTTP path to respond to.
# Can also be set with the environment variable SIDEKIQ_ALIVE_PATH.
# default: '/'
#
# config.path = '/'
# ==> Custom Liveness Probe
# Extra check to decide if restart the pod or not for example connection to DB.
# `false`, `nil` or `raise` will not write the liveness probe
# default: proc { true }
#
# config.custom_liveness_probe = proc { db_running? }
# ==> Liveness key
# Key to be stored in Redis as probe of liveness
# default: "SIDEKIQ::LIVENESS_PROBE_TIMESTAMP"
#
# config.liveness_key = "SIDEKIQ::LIVENESS_PROBE_TIMESTAMP"
# ==> Time to live
# Time for the key to be kept by Redis.
# Here is where you can set de periodicity that the Sidekiq has to probe it is working
# Time unit: seconds
# default: 10 * 60 # 10 minutes
#
# config.time_to_live = 10 * 60
# ==> Callback
# After the key is stored in redis you can perform anything.
# For example a webhook or email to notify the team
# default: proc {}
#
# require 'net/http'
# config.callback = proc { Net::HTTP.get("https://status.com/ping") }
# ==> Shutdown callback
# When sidekiq process is shutting down, you can perform some arbitrary action.
# default: proc {}
#
# config.shutdown_callback = proc { puts "Sidekiq is shutting down" }
# ==> Queue Prefix
# SidekiqAlive will run in a independent queue for each instance/replica
# This queue name will be generated with: "#{queue_prefix}-#{hostname}.
# You can customize the prefix here.
# default: :sidekiq-alive
#
# config.queue_prefix = :other
# ==> Concurrency
# The maximum number of Redis connections requested for the SidekiqAlive pool.
# Can also be set with the environment variable SIDEKIQ_ALIVE_CONCURRENCY.
# NOTE: only effects Sidekiq 7 or greater.
# default: 2
#
# config.concurrency = 3
# ==> Rack server
# Web server used to serve an HTTP response. By default simple GServer based http server is used.
# To use specific server, rack gem version > 2 is required. For rack version >= 3, rackup gem is required.
# Can also be set with the environment variable SIDEKIQ_ALIVE_SERVER.
# default: nil
#
# config.server = 'puma'
# ==> Quiet mode timeout in seconds
# When sidekiq is shutting down, the Sidekiq process stops pulling jobs from the queue. This includes alive key update job. In case of
# long running jobs, alive key can expire before the job is finished. To avoid this, web server is set in to quiet mode
# and is returning 200 OK for healthcheck requests. To avoid infinite quiet mode in case sidekiq process is stuck in shutdown,
# timeout can be set. After timeout is reached, web server resumes normal operations and will return unhealthy status in case
# alive key is expired or purged from redis.
# default: 180
#
# config.quiet_timeout = 300
end
After checking out the repo, run bin/setup
to install dependencies. Then, run rake spec
to run the tests. You can also run bin/console
for an interactive prompt that will allow you to experiment.
To install this gem onto your local machine, run bundle exec rake install
.
Here is an example rails app
Bug reports and pull requests are welcome on GitHub at https://github.com/arturictus/sidekiq_alive. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the Contributor Covenant code of conduct.
The gem is available as open source under the terms of the MIT License.