debian-mirror-gitlab/lib/gitlab/database/load_balancing/load_balancer.rb

278 lines
9.3 KiB
Ruby
Raw Normal View History

2021-09-04 01:27:46 +05:30
# frozen_string_literal: true
module Gitlab
module Database
module LoadBalancing
# Load balancing for ActiveRecord connections.
#
# Each host in the load balancer uses the same credentials as the primary
# database.
class LoadBalancer
CACHE_KEY = :gitlab_load_balancer_host
2021-10-27 15:23:28 +05:30
REPLICA_SUFFIX = '_replica'
2021-09-04 01:27:46 +05:30
2021-11-18 22:05:49 +05:30
attr_reader :name, :host_list, :configuration
2021-11-11 11:23:49 +05:30
# configuration - An instance of `LoadBalancing::Configuration` that
# contains the configuration details (such as the hosts)
# for this load balancer.
2021-11-18 22:05:49 +05:30
def initialize(configuration)
2021-11-11 11:23:49 +05:30
@configuration = configuration
2021-11-18 22:05:49 +05:30
@primary_only = !configuration.load_balancing_enabled?
2021-11-11 11:23:49 +05:30
@host_list =
2021-11-18 22:05:49 +05:30
if @primary_only
2021-11-11 11:23:49 +05:30
HostList.new([PrimaryHost.new(self)])
else
HostList.new(configuration.hosts.map { |addr| Host.new(addr, self) })
end
2021-11-18 22:05:49 +05:30
@name = @configuration.model.connection_db_config.name.to_sym
end
def primary_only?
@primary_only
2021-10-27 15:23:28 +05:30
end
def disconnect!(timeout: 120)
host_list.hosts.each { |host| host.disconnect!(timeout: timeout) }
2021-09-04 01:27:46 +05:30
end
# Yields a connection that can be used for reads.
#
# If no secondaries were available this method will use the primary
# instead.
def read(&block)
conflict_retried = 0
while host
ensure_caching!
begin
connection = host.connection
return yield connection
rescue StandardError => error
2021-11-18 22:05:49 +05:30
if primary_only?
# If we only have primary configured, retrying is pointless
raise error
elsif serialization_failure?(error)
2021-09-04 01:27:46 +05:30
# This error can occur when a query conflicts. See
# https://www.postgresql.org/docs/current/static/hot-standby.html#HOT-STANDBY-CONFLICT
# for more information.
#
# In this event we'll cycle through the secondaries at most 3
# times before using the primary instead.
will_retry = conflict_retried < @host_list.length * 3
LoadBalancing::Logger.warn(
event: :host_query_conflict,
message: 'Query conflict on host',
conflict_retried: conflict_retried,
will_retry: will_retry,
db_host: host.host,
db_port: host.port,
host_list_length: @host_list.length
)
if will_retry
conflict_retried += 1
release_host
else
break
end
elsif connection_error?(error)
host.offline!
release_host
else
raise error
end
end
end
LoadBalancing::Logger.warn(
event: :no_secondaries_available,
message: 'No secondaries were available, using primary instead',
conflict_retried: conflict_retried,
host_list_length: @host_list.length
)
read_write(&block)
end
# Yields a connection that can be used for both reads and writes.
def read_write
connection = nil
# In the event of a failover the primary may be briefly unavailable.
# Instead of immediately grinding to a halt we'll retry the operation
# a few times.
retry_with_backoff do
2021-10-27 15:23:28 +05:30
connection = pool.connection
2021-09-04 01:27:46 +05:30
yield connection
end
end
# Returns a host to use for queries.
#
# Hosts are scoped per thread so that multiple threads don't
# accidentally re-use the same host + connection.
def host
2021-10-27 15:23:28 +05:30
request_cache[CACHE_KEY] ||= @host_list.next
2021-09-04 01:27:46 +05:30
end
# Releases the host and connection for the current thread.
def release_host
2021-10-27 15:23:28 +05:30
if host = request_cache[CACHE_KEY]
2021-09-04 01:27:46 +05:30
host.disable_query_cache!
host.release_connection
end
2021-10-27 15:23:28 +05:30
request_cache.delete(CACHE_KEY)
2021-09-04 01:27:46 +05:30
end
def release_primary_connection
2021-10-27 15:23:28 +05:30
pool.release_connection
2021-09-04 01:27:46 +05:30
end
# Returns the transaction write location of the primary.
def primary_write_location
location = read_write do |connection|
2021-10-27 15:23:28 +05:30
::Gitlab::Database.main.get_write_location(connection)
2021-09-04 01:27:46 +05:30
end
return location if location
raise 'Failed to determine the write location of the primary database'
end
# Returns true if there was at least one host that has caught up with the given transaction.
def select_up_to_date_host(location)
all_hosts = @host_list.hosts.shuffle
host = all_hosts.find { |host| host.caught_up?(location) }
return false unless host
2021-10-27 15:23:28 +05:30
request_cache[CACHE_KEY] = host
2021-09-04 01:27:46 +05:30
true
end
# Yields a block, retrying it upon error using an exponential backoff.
def retry_with_backoff(retries = 3, time = 2)
2021-11-18 22:05:49 +05:30
# In CI we only use the primary, but databases may not always be
# available (or take a few seconds to become available). Retrying in
# this case can slow down CI jobs. In addition, retrying with _only_
# a primary being present isn't all that helpful.
#
# To prevent this from happening, we don't make any attempt at
# retrying unless one or more replicas are used. This matches the
# behaviour from before we enabled load balancing code even if no
# replicas were configured.
return yield if primary_only?
2021-09-04 01:27:46 +05:30
retried = 0
last_error = nil
while retried < retries
begin
return yield
rescue StandardError => error
raise error unless connection_error?(error)
# We need to release the primary connection as otherwise Rails
# will keep raising errors when using the connection.
release_primary_connection
last_error = error
sleep(time)
retried += 1
time **= 2
end
end
raise last_error
end
def connection_error?(error)
case error
2021-11-18 22:05:49 +05:30
when ActiveRecord::NoDatabaseError
# Retrying this error isn't going to magically make the database
# appear. It also slows down CI jobs that are meant to create the
# database in the first place.
false
2021-09-04 01:27:46 +05:30
when ActiveRecord::StatementInvalid, ActionView::Template::Error
# After connecting to the DB Rails will wrap query errors using this
# class.
2021-11-11 11:23:49 +05:30
if (cause = error.cause)
connection_error?(cause)
else
false
end
2021-09-04 01:27:46 +05:30
when *CONNECTION_ERRORS
true
else
# When PG tries to set the client encoding but fails due to a
# connection error it will raise a PG::Error instance. Catching that
# would catch all errors (even those we don't want), so instead we
# check for the message of the error.
error.message.start_with?('invalid encoding name:')
end
end
def serialization_failure?(error)
if error.cause
serialization_failure?(error.cause)
else
error.is_a?(PG::TRSerializationFailure)
end
end
2021-10-27 15:23:28 +05:30
# pool_size - The size of the DB pool.
# host - An optional host name to use instead of the default one.
# port - An optional port to connect to.
def create_replica_connection_pool(pool_size, host = nil, port = nil)
db_config = pool.db_config
2021-09-04 01:27:46 +05:30
2021-10-27 15:23:28 +05:30
env_config = db_config.configuration_hash.dup
env_config[:pool] = pool_size
env_config[:host] = host if host
env_config[:port] = port if port
replica_db_config = ActiveRecord::DatabaseConfigurations::HashConfig.new(
db_config.env_name,
db_config.name + REPLICA_SUFFIX,
env_config
)
2021-09-04 01:27:46 +05:30
2021-10-27 15:23:28 +05:30
# We cannot use ActiveRecord::Base.connection_handler.establish_connection
# as it will rewrite ActiveRecord::Base.connection
ActiveRecord::ConnectionAdapters::ConnectionHandler
.new
.establish_connection(replica_db_config)
2021-09-04 01:27:46 +05:30
end
2021-10-27 15:23:28 +05:30
# ActiveRecord::ConnectionAdapters::ConnectionHandler handles fetching,
# and caching for connections pools for each "connection", so we
# leverage that.
def pool
ActiveRecord::Base.connection_handler.retrieve_connection_pool(
2021-11-11 11:23:49 +05:30
@configuration.model.connection_specification_name,
2021-10-27 15:23:28 +05:30
role: ActiveRecord::Base.writing_role,
shard: ActiveRecord::Base.default_shard
2021-11-18 22:05:49 +05:30
) || raise(::ActiveRecord::ConnectionNotEstablished)
2021-10-27 15:23:28 +05:30
end
2021-11-11 11:23:49 +05:30
private
2021-10-27 15:23:28 +05:30
def ensure_caching!
host.enable_query_cache! unless host.query_cache_enabled
2021-09-04 01:27:46 +05:30
end
2021-10-27 15:23:28 +05:30
def request_cache
2021-11-11 11:23:49 +05:30
base = SafeRequestStore[:gitlab_load_balancer] ||= {}
base[self] ||= {}
2021-09-04 01:27:46 +05:30
end
end
end
end
end