require 'spec_helper' require 'securerandom' require 'redis' require 'connection_pool' RSpec.describe Redlock::Client do # It is recommended to have at least 3 servers in production let(:lock_manager_opts) { { retry_count: 3 } } let(:lock_manager) { Redlock::Client.new(Redlock::Client::DEFAULT_REDIS_URLS, lock_manager_opts) } let(:redis_client) { Redis.new(url: "redis://#{redis1_host}:#{redis1_port}") } let(:resource_key) { SecureRandom.hex(3) } let(:ttl) { 1000 } let(:redis1_host) { ENV["REDIS1_HOST"] || "localhost" } let(:redis1_port) { ENV["REDIS1_PORT"] || "6379" } let(:redis2_host) { ENV["REDIS2_HOST"] || "127.0.0.1" } let(:redis2_port) { ENV["REDIS2_PORT"] || "6379" } let(:redis3_host) { ENV["REDIS3_HOST"] || "127.0.0.1" } let(:redis3_port) { ENV["REDIS3_PORT"] || "6379" } let(:unreachable_redis) { redis = Redis.new(url: 'redis://localhost:46864') def redis.with yield self end redis } describe 'initialize' do it 'accepts both redis URLs and Redis objects' do servers = [ "redis://#{redis1_host}:#{redis1_port}", Redis.new(url: "redis://#{redis2_host}:#{redis2_port}") ] redlock = Redlock::Client.new(servers) redlock_servers = redlock.instance_variable_get(:@servers).map do |s| s.instance_variable_get(:@redis).connection[:host] end expect(redlock_servers).to match_array([redis1_host, redis2_host]) end it 'accepts ConnectionPool objects' do pool = ConnectionPool.new { Redis.new(url: "redis://#{redis1_host}:#{redis1_port}") } redlock = Redlock::Client.new([pool]) lock_info = lock_manager.lock(resource_key, ttl) expect(resource_key).to_not be_lockable(lock_manager, ttl) lock_manager.unlock(lock_info) end end describe 'lock' do context 'when lock is available' do after(:each) { lock_manager.unlock(@lock_info) if @lock_info } it 'locks' do @lock_info = lock_manager.lock(resource_key, ttl) expect(resource_key).to_not be_lockable(lock_manager, ttl) end it 'returns lock information' do @lock_info = lock_manager.lock(resource_key, ttl) expect(@lock_info).to be_lock_info_for(resource_key) end it 'interprets lock time as milliseconds' do ttl = 20000 @lock_info = lock_manager.lock(resource_key, ttl) expect(redis_client.pttl(resource_key)).to be_within(200).of(ttl) end it 'can extend its own lock' do my_lock_info = lock_manager.lock(resource_key, ttl) @lock_info = lock_manager.lock(resource_key, ttl, extend: my_lock_info) expect(@lock_info).to be_lock_info_for(resource_key) expect(@lock_info[:value]).to eq(my_lock_info[:value]) end context 'when extend param is nil' do it 'defaults to creating a new lock' do @lock_info = lock_manager.lock(resource_key, ttl, extend: nil) expect(@lock_info).to be_lock_info_for(resource_key) expect(@lock_info[:value]).to be end end context 'when extend_only_if_locked flag is given' do it 'does not extend a non-existent lock' do @lock_info = lock_manager.lock(resource_key, ttl, extend: {value: 'hello world'}, extend_only_if_locked: true) expect(@lock_info).to eq(false) end end it '(when extending) resets the TTL, rather than adding extra time to it' do ttl = 20000 lock_info = lock_manager.lock(resource_key, ttl) expect(resource_key).to_not be_lockable(lock_manager, ttl) lock_info = lock_manager.lock(resource_key, ttl, extend: lock_info, extend_only_if_locked: true) expect(lock_info).not_to be_nil expect(redis_client.pttl(resource_key)).to be_within(200).of(ttl) end context 'when extend_only_if_locked flag is not given' do it "sets the given value when trying to extend a non-existent lock" do @lock_info = lock_manager.lock(resource_key, ttl, extend: {value: 'hello world'}, extend_only_if_locked: false) expect(@lock_info).to be_lock_info_for(resource_key) expect(@lock_info[:value]).to eq('hello world') # really we should test what's in redis end end it "doesn't extend somebody else's lock" do @lock_info = lock_manager.lock(resource_key, ttl) second_attempt = lock_manager.lock(resource_key, ttl) expect(second_attempt).to eq(false) end context 'when extend_life flag is given' do it 'treats it as extend_only_if_locked but warns it is deprecated' do ttl = 20_000 lock_info = lock_manager.lock(resource_key, ttl) expect(resource_key).to_not be_lockable(lock_manager, ttl) expect(lock_manager).to receive(:warn).with(/DEPRECATION WARNING: The `extend_life`/) lock_info = lock_manager.lock(resource_key, ttl, extend: lock_info, extend_life: true) expect(lock_info).not_to be_nil end end context 'when extend_only_if_life flag is given' do it 'treats it as extend_only_if_locked but warns it is deprecated' do ttl = 20_000 lock_info = lock_manager.lock(resource_key, ttl) expect(resource_key).to_not be_lockable(lock_manager, ttl) expect(lock_manager).to receive(:warn).with(/DEPRECATION WARNING: The `extend_only_if_life`/) lock_info = lock_manager.lock(resource_key, ttl, extend: lock_info, extend_only_if_life: true) expect(lock_info).not_to be_nil end end end context 'when lock is not available' do before { @another_lock_info = lock_manager.lock(resource_key, ttl) } after { lock_manager.unlock(@another_lock_info) } it 'returns false' do lock_info = lock_manager.lock(resource_key, ttl) expect(lock_info).to eql(false) end it "can't extend somebody else's lock" do yet_another_lock_info = @another_lock_info.merge value: 'gibberish' lock_info = lock_manager.lock(resource_key, ttl, extend: yet_another_lock_info) expect(lock_info).to eql(false) end it 'tries up to \'retry_count\' + 1 times' do expect(lock_manager).to receive(:lock_instances).exactly( lock_manager_opts[:retry_count] + 1).times.and_return(false) lock_manager.lock(resource_key, ttl) end it 'sleeps in between retries' do expect(lock_manager).to receive(:sleep).exactly(lock_manager_opts[:retry_count]).times lock_manager.lock(resource_key, ttl) end it 'sleeps at least the specified retry_delay in milliseconds' do expected_minimum = described_class::DEFAULT_RETRY_DELAY expect(lock_manager).to receive(:sleep) do |sleep| expect(sleep).to satisfy { |value| value >= expected_minimum / 1000.to_f } end.at_least(:once) lock_manager.lock(resource_key, ttl) end it 'sleeps a maximum of retry_delay + retry_jitter in milliseconds' do expected_maximum = described_class::DEFAULT_RETRY_DELAY + described_class::DEFAULT_RETRY_JITTER expect(lock_manager).to receive(:sleep) do |sleep| expect(sleep).to satisfy { |value| value < expected_maximum / 1000.to_f } end.at_least(:once) lock_manager.lock(resource_key, ttl) end it 'accepts retry_delay as proc' do retry_delay = proc do |attempt_number| expect(attempt_number).to eq(1) 2000 end lock_manager = Redlock::Client.new(Redlock::Client::DEFAULT_REDIS_URLS, retry_count: 1, retry_delay: retry_delay) another_lock_info = lock_manager.lock(resource_key, ttl) expect(lock_manager).to receive(:sleep) do |sleep| expect(sleep * 1000).to be_within(described_class::DEFAULT_RETRY_JITTER).of(2000) end.exactly(:once) lock_manager.lock(resource_key, ttl) lock_manager.unlock(another_lock_info) end end context 'when a server goes away' do it 'does not raise an error on connection issues' do # We re-route the lock manager to a (hopefully) non-existent Redis URL. redis_instance = lock_manager.instance_variable_get(:@servers).first redis_instance.instance_variable_set(:@redis, unreachable_redis) expect { expect(lock_manager.lock(resource_key, ttl)).to be_falsey }.to_not raise_error end end context 'when a server comes back' do it 'recovers from connection issues' do # Same as above. redis_instance = lock_manager.instance_variable_get(:@servers).first old_redis = redis_instance.instance_variable_get(:@redis) redis_instance.instance_variable_set(:@redis, unreachable_redis) expect(lock_manager.lock(resource_key, ttl)).to be_falsey redis_instance.instance_variable_set(:@redis, old_redis) expect(lock_manager.lock(resource_key, ttl)).to be_truthy end end context 'when script cache has been flushed' do before(:each) do @manipulated_instance = lock_manager.instance_variable_get(:@servers).first @manipulated_instance.instance_variable_get(:@redis).script(:flush) end it 'does not raise a Redis::CommandError: NOSCRIPT error' do expect { lock_manager.lock(resource_key, ttl) }.to_not raise_error end it 'tries to load the scripts to cache again' do expect(@manipulated_instance).to receive(:load_scripts).and_call_original lock_manager.lock(resource_key, ttl) end context 'when the script re-loading fails' do it 'does not try to to load the scripts to cache again twice' do # This time we do not pass it through to Redis, in order to simulate a passing # call to LOAD SCRIPT followed by another NOSCRIPT error. Imagine someone # repeatedly calling SCRIPT FLUSH on our Redis instance. expect(@manipulated_instance).to receive(:load_scripts) expect { lock_manager.lock(resource_key, ttl) }.to raise_error(/NOSCRIPT/) end end context 'when the script re-loading succeeds' do it 'locks' do expect(lock_manager.lock(resource_key, ttl)).to be_lock_info_for(resource_key) end end end describe 'block syntax' do context 'when lock is available' do it 'locks' do lock_manager.lock(resource_key, ttl) do |_| expect(resource_key).to_not be_lockable(lock_manager, ttl) end end it 'passes lock information as block argument' do lock_manager.lock(resource_key, ttl) do |lock_info| expect(lock_info).to be_lock_info_for(resource_key) end end it 'returns true' do rv = lock_manager.lock(resource_key, ttl) {} expect(rv).to eql(true) end it 'automatically unlocks' do lock_manager.lock(resource_key, ttl) {} expect(resource_key).to be_lockable(lock_manager, ttl) end it 'automatically unlocks when block raises exception' do lock_manager.lock(resource_key, ttl) { fail } rescue nil expect(resource_key).to be_lockable(lock_manager, ttl) end end context 'when lock is not available' do before { @another_lock_info = lock_manager.lock(resource_key, ttl) } after { lock_manager.unlock(@another_lock_info) } it 'passes false as block argument' do lock_manager.lock(resource_key, ttl) do |lock_info| expect(lock_info).to eql(false) end end it 'returns false' do rv = lock_manager.lock(resource_key, ttl) {} expect(rv).to eql(false) end end end end describe 'unlock' do before { @lock_info = lock_manager.lock(resource_key, ttl) } it 'unlocks' do expect(resource_key).to_not be_lockable(lock_manager, ttl) lock_manager.unlock(@lock_info) expect(resource_key).to be_lockable(lock_manager, ttl) end end describe 'lock!' do context 'when lock is available' do it 'locks' do lock_manager.lock!(resource_key, ttl) do expect(resource_key).to_not be_lockable(lock_manager, ttl) end end it "returns the received block's return value" do rv = lock_manager.lock!(resource_key, ttl) { :success } expect(rv).to eql(:success) end it 'automatically unlocks' do lock_manager.lock!(resource_key, ttl) {} expect(resource_key).to be_lockable(lock_manager, ttl) end it 'automatically unlocks when block raises exception' do lock_manager.lock!(resource_key, ttl) { fail } rescue nil expect(resource_key).to be_lockable(lock_manager, ttl) end it 'passes the extension parameter' do my_lock_info = lock_manager.lock(resource_key, ttl) expect{ lock_manager.lock!(resource_key, ttl, extend: my_lock_info){} }.to_not raise_error end end context 'when lock is not available' do before { @another_lock_info = lock_manager.lock(resource_key, ttl) } after { lock_manager.unlock(@another_lock_info) } it 'raises a LockError' do expect { lock_manager.lock!(resource_key, ttl) {} }.to raise_error( Redlock::LockError, "failed to acquire lock on '#{resource_key}'" ) end it 'does not execute the block' do expect do begin lock_manager.lock!(resource_key, ttl) { fail } rescue Redlock::LockError end end.to_not raise_error end end end describe 'get_remaining_ttl_for_resource' do context 'when lock is valid' do after(:each) { lock_manager.unlock(@lock_info) if @lock_info } it 'gets the remaining ttl of a lock' do ttl = 20_000 @lock_info = lock_manager.lock(resource_key, ttl) remaining_ttl = lock_manager.get_remaining_ttl_for_resource(resource_key) expect(remaining_ttl).to be_within(300).of(ttl) end context 'when servers respond with varying ttls' do let (:servers) { [ "redis://#{redis1_host}:#{redis1_port}", "redis://#{redis2_host}:#{redis2_port}", "redis://#{redis3_host}:#{redis3_port}" ] } let (:redlock) { Redlock::Client.new(servers) } after(:each) { redlock.unlock(@lock_info) if @lock_info } it 'returns the minimum ttl value' do ttl = 20_000 @lock_info = redlock.lock(resource_key, ttl) # Mock redis server responses to return different ttls returned_ttls = [20_000, 15_000, 10_000] redlock.instance_variable_get(:@servers).each_with_index do |server, index| allow(server).to(receive(:get_remaining_ttl)) .with(resource_key) .and_return([@lock_info[:value], returned_ttls[index]]) end remaining_ttl = redlock.get_remaining_ttl_for_lock(@lock_info) # Assert that the TTL is closest to the closest to the correct value expect(remaining_ttl).to be_within(300).of(returned_ttls[1]) end end end context 'when lock is not valid' do it 'returns nil' do lock_info = lock_manager.lock(resource_key, ttl) lock_manager.unlock(lock_info) remaining_ttl = lock_manager.get_remaining_ttl_for_resource(resource_key) expect(remaining_ttl).to be_nil end end context 'when server goes away' do after(:each) { lock_manager.unlock(@lock_info) if @lock_info } it 'does not raise an error on connection issues' do @lock_info = lock_manager.lock(resource_key, ttl) # Replace redis with unreachable instance redis_instance = lock_manager.instance_variable_get(:@servers).first old_redis = redis_instance.instance_variable_get(:@redis) redis_instance.instance_variable_set(:@redis, unreachable_redis) expect { remaining_ttl = lock_manager.get_remaining_ttl_for_resource(resource_key) expect(remaining_ttl).to be_nil }.to_not raise_error end end context 'when a server comes back' do after(:each) { lock_manager.unlock(@lock_info) if @lock_info } it 'recovers from connection issues' do @lock_info = lock_manager.lock(resource_key, ttl) # Replace redis with unreachable instance redis_instance = lock_manager.instance_variable_get(:@servers).first old_redis = redis_instance.instance_variable_get(:@redis) redis_instance.instance_variable_set(:@redis, unreachable_redis) expect(lock_manager.get_remaining_ttl_for_resource(resource_key)).to be_nil # Restore redis redis_instance.instance_variable_set(:@redis, old_redis) expect(lock_manager.get_remaining_ttl_for_resource(resource_key)).to be_truthy end end end describe 'get_remaining_ttl_for_lock' do context 'when lock is valid' do it 'gets the remaining ttl of a lock' do ttl = 20_000 lock_info = lock_manager.lock(resource_key, ttl) remaining_ttl = lock_manager.get_remaining_ttl_for_lock(lock_info) expect(remaining_ttl).to be_within(300).of(ttl) lock_manager.unlock(lock_info) end end context 'when lock is not valid' do it 'returns nil' do lock_info = lock_manager.lock(resource_key, ttl) lock_manager.unlock(lock_info) remaining_ttl = lock_manager.get_remaining_ttl_for_lock(lock_info) expect(remaining_ttl).to be_nil end end end describe 'locked?' do context 'when lock is available' do after(:each) { lock_manager.unlock(@lock_info) if @lock_info } it 'returns true' do @lock_info = lock_manager.lock(resource_key, ttl) expect(lock_manager).to be_locked(resource_key) end end context 'when lock is not available' do it 'returns false' do lock_info = lock_manager.lock(resource_key, ttl) lock_manager.unlock(lock_info) expect(lock_manager).not_to be_locked(resource_key) end end end describe 'valid_lock?' do context 'when lock is available' do after(:each) { lock_manager.unlock(@lock_info) if @lock_info } it 'returns true' do @lock_info = lock_manager.lock(resource_key, ttl) expect(lock_manager).to be_valid_lock(@lock_info) end end context 'when lock is not available' do it 'returns false' do lock_info = lock_manager.lock(resource_key, ttl) lock_manager.unlock(lock_info) expect(lock_manager).not_to be_valid_lock(lock_info) end end end describe '#default_time_source' do context 'when CLOCK_MONOTONIC is available (MRI, JRuby)' do it 'returns a callable using Process.clock_gettime()' do skip 'CLOCK_MONOTONIC not defined' unless defined?(Process::CLOCK_MONOTONIC) expect(Process).to receive(:clock_gettime).with(Process::CLOCK_MONOTONIC).and_call_original Redlock::Client.default_time_source.call() end end context 'when CLOCK_MONOTONIC is not available' do it 'returns a callable using Time.now()' do cm = Process.send(:remove_const, :CLOCK_MONOTONIC) expect(Time).to receive(:now).and_call_original Redlock::Client.default_time_source.call() Process.const_set(:CLOCK_MONOTONIC, cm) if cm end end end end