Key based pessimistic locking for Ruby and MongoDB. Is this key avaliable? Yes - Lock it for me for a sec will you. No - OK I'll just wait here until its ready.
It handles timeouts and and vanishing lock owners (such as machine failures)
Add this line to your application's Gemfile:
gem 'mongo-lock'
And then execute:
$ bundle
Or install it yourself as:
$ gem install mongo-lock
Build your indexes on any collection that is going to hold locks:
Mongo::Lock.ensure_indexes # Will use the collection provided to #configure
For this to work you must have configured your collection or collections in when intializing locks, in #configure or .configure.
We took quite a bit of inspiration from the redis versions of this gem by @mlanett and @PatrickTulskie. If you aren't already using MongoDB or are already using Redis in your stack you probably want to think about using one of them.
We also looked at mongo-locking gem by @servio. It was a bit complicated for ours needs, but if you need to lock related and embedded documents rather than just keys it could be what you need.
A lock has an expected lifetime. If the owner of a lock disappears (due to machine failure, network failure, process death), you want the lock to expire and another owner to be able to acquire the lock. At the same time, the owner of a lock should be able to extend its lifetime. Thus, you can acquire a lock with a conservative estimate on lifetime, and extend it as necessary, rather than acquiring the lock with a very long lifetime which will result in long waits in the event of failures.
A lock has an owner. Mongo::Lock defaults to using an owner id of HOSTNAME:PID:TID.
Mongo::Lock makes no effort to help configure the MongoDB connection - that's what the Mongo driver is for, you can use either Moped or the Mongo Ruby Driver. If you are using Mongoid'll you want to be using the Moped driver. Mongo::Lock will automatically choose the right driver for the collection you provide and raise an error if you try and mix them.
Mongo::Lock.configure collection: Mongo::Connection.new("localhost").db("somedb").collection("locks")
Or using Moped:
session = Moped::Session.new([ "127.0.0.1:27017" ])
session.use 'locks_db'
Mongo::Lock.configure collection: session[:locks]
Or if your session is already set up with Mongoid:
Mongo::Lock.configure collection: Mongoid.session(:default)[:locks]
You can add multiple collections with a hash that can be referenced later using symbols:
Mongo::Lock.configure collections: { default: database.collection("locks"), other: database.collection("other_locks") }
Mongo::Lock.acquire('my_lock') # Locks in the default collection
Mongo::Lock.acquire('my_lock', collection: :other) # Locks in the other_locks collection
You can also configure using a block:
Mongo::Lock.configure do |config|
config.collections: {
default: database.collection("locks"),
other: database.collection("other_locks")
}
end
A lock may need more than one attempt to acquire it. Mongo::Lock offers:
Mongo::Lock.configure do |config|
config.timeout_in = false # timeout_in in seconds on acquisition; this defaults to false ie no time limit.
config.limit = 100 # The limit on the number of acquisition attempts; this defaults to 100.
config.frequency = 1 # Frequency in seconds for acquisition attempts ; this defaults to 1.
# acquisition_attempt_frequency can also be given as a proc which will be passed the attempt number
config.frequency = Proc.new { |x| x**2 }
end
A lock will automatically be relinquished once its expiry has passed. Expired locks are cleaned up by MongoDB's TTL index, which may take up to 60 seconds or more depending on load to actually remove expired locks. Expired locks that have not been cleaned out can still be acquire. You must have built your indexes to ensure expired locks are cleaned out.
Mongo::Lock.configure do |config|
config.expire_in = false # timeout_in in seconds for lock expiry; this defaults to 10.
end
You can remove expired locks yourself with:
Mongo::Lock.clean_expired
If a lock cannot be acquired, released or extended it will return false, you can set the raise option to true to raise a Mongo::Lock::LockNotAcquiredError or Mongo::Lock::LockNotReleasedError.
Mongo::Lock.configure do |config|
config.should_raise = true # Whether to raise an error when acquire, release or extend fail.
end
Using .acquire!, #acquire!, .release!, #release!, #extend_by! and #extend! will also raise exceptions instead of returning false.
By default the owner id will be generated using the following Proc:
Proc.new { "#{`hostname`.strip}:#{Process.pid}:#{Thread.object_id}" }
You can override this with either a Proc that returns any object that responds to to_s, or with any object that responds to #to_s.
Mongo::Lock.configure do |config|
config.owner = ['my', 'owner', 'id']
end
# Or
Mongo::Lock.configure do |config|
config.owner = Proc.new { [`hostname`.strip, Process.pid] }
end
Note: Hosts, threads or processes using the same owner can acquire each others locks.
You can use Mongo::Lock's class methods:
Mongo::Lock.acquire('my_key', options) do |lock|
# Do Something here that needs my_key locked
end
lock = Mongo::Lock.new('my_key', options).acquire
# Do Something here that needs my_key locked
lock.release
# or
Mongo::Lock.release('my_key')
Or you can initialise your own instance.
Mongo::Lock.new('my_key', options).acquire do |lock|
# Do Something here that needs my_key locked
end
lock = Mongo::Lock.acquire('my_key', options)
# Do Something here that needs my_key locked
lock.release
# or
Mongo::Lock.release('my_key')
The lock key is treated in the same way as ActiveSupport::Cache's keys, except instead of responding to :cache_key or to :to_param it should respond to :lock_key or to :to_param. You can use Hashes and Arrays of values as cache keys.
When using Mongo::Lock#acquire, Mongo::Lock#release or Mongo::Lock#new after the key you may overide any of the following options:
Mongo::Lock.new 'my_key', {
collection: Mongo::Connection.new("localhost").db("somedb").collection("locks"), # May also be a symbol if that symbol was provided in the collections hash to Mongo::Lock.configure
timeout_in: 10, # timeout_in in seconds on acquisition; this defaults to false ie no time limit.
limit: 10, # The limit on the number of acquisition attempts; this defaults to 100.
frequency: 2, # Frequency in seconds for acquisition attempts ; this defaults to 1.
expire_in: 10,# timeout_in in seconds for lock expiry; this defaults to 10.
}
You can extend a lock by calling Mongo::Lock#extend_by with the number of seconds to extend the lock.
Mongo::Lock.new 'my_key' do |lock|
lock.extend_by 10
end
You can also call Mongo::Lock#extend and it will extend by the lock's expire_in option.
Mongo::Lock.new 'my_key' do |lock|
lock.extend
end
Mongo::Lock.acquire 'my_key', expire_in: 10 do |lock|
sleep 9
lock.expired? # False
sleep 11
lock.expired? # True
end
Mongo::Lock.available? 'my_key'
# Or
lock = Mongo::Lock.new('my_key')
lock.available?
You can release all locks across an entire collection or owner with the .release_all method.
Mongo::Lock.release_all # Release all locks in all registered collections
Mongo::Lock.release_all collection: :my_locks # Release all locks in the collection registered as :my_locks
Mongo::Lock.release_all collection: my_collection # Release all locks in this instance of Mongo::Collection
Mongo::Lock.release_all collections: [c1,c2] # Release all locks in these instances of Mongo::Collection
Mongo::Lock.release_all collections: {a: ca, b: cb} # Release all locks in these instances of Mongo::Collection
Mongo::Lock.release_all owner: 'me' # Release all locks in all registered collections that belong to 'me'
You can clear expire locks from the database with the .clear_expired method. If you have called .ensure_indexes mongo will do this for you automatically with a time to live index.
Mongo::Lock.clear_expired # Clear expired locks in all registered collections
Mongo::Lock.clear_expired collection: :my_locks # Clear expired locks in the collection registered as :my_locks
Mongo::Lock.clear_expired collection: my_collection # Clear expired locks in this instance of Mongo::Collection
Mongo::Lock.clear_expired collections: [c1,c2] # Clear expired locks in these instances of Mongo::Collection
Mongo::Lock.clear_expired collections: {a: ca, b: cb} # Clear expired locks in these instances of Mongo::Collection
Mongo::Lock.available? 'my_key'
# Or
lock = Mongo::Lock.new('my_key')
lock.available?
If Mongo::Lock#acquire cannot acquire a lock within its configuration limits it will return false.
unless Mongo::Lock.acquire 'my_key'
# Maybe try again tomorrow
end
If Mongo::Lock#release cannot release a lock because it wasn't acquired it will return false. If it has already been released, or has expired it will do nothing and return true.
unless Mongo::Lock.release 'my_key'
# Eh somebody else should release it eventually
end
If Mongo::Lock#extend cannot be extended because it has already been released, it is owned by someone else or it was never acquired it will return false.
unless lock.extend_by 10
# Eh somebody else should release it eventually
end
If the should_raise error option is set to true or you append ! to the end of the method name and you call any of the acquire, release, extend_by or extend methods they will raise a Mongo::Lock::NotAcquiredError, Mongo::Lock::NotReleasedError or Mongo::Lock::NotExtendedError instead of returning false.
begin
Mongo::Lock.acquire! 'my_key'
rescue Mongo::Lock::LockNotAcquiredError => e
# Maybe try again tomorrow
end
# Or
begin
Mongo::Lock.acquire 'my_key', should\_raise: true
rescue Mongo::Lock::LockNotAcquiredError => e
# Maybe try again tomorrow
end
If you are running mongo-lock inside Rails it will add the following rake tasks for you.
bundle exec rake mongolock:clear_expired # Calls Mongo::Lock.clear_expired
bundle exec rake mongolock:release_all # Calls Mongo::Lock.release_all
bundle exec rake mongolock:ensure_indexes # Calls Mongo::Lock.ensure_indexes
Matthew Spence (msaspence)
The bulk of this gem has been developed for and by trak.io
- Fork it
- Create your feature branch (
git checkout -b my-new-feature
) - Commit your changes (
git commit -am 'Added some feature'
) - Push to the branch (
git push origin my-new-feature
) - Create new Pull Request