@ExperimentalApi(value="https://github.com/grpc/grpc-java/issues/1771") @NotThreadSafe public abstract class LoadBalancer extends Object
NameResolver
and provides the
channel a usable subchannel when asked.
A LoadBalancer typically implements three interfaces:
LoadBalancer
is the main interface. All methods on it are invoked sequentially
in the same synchronization context (see next section) as returned by
LoadBalancer.Helper.getSynchronizationContext()
. It receives the results
from the NameResolver
, updates of subchannels' connectivity states, and the
channel's request for the LoadBalancer to shutdown.SubchannelPicker
does the actual load-balancing work. It selects
a Subchannel
for each new RPC.Factory
creates a new LoadBalancer
instance.
Helper
is implemented by gRPC library and provided to Factory
. It provides functionalities that a LoadBalancer
implementation would typically
need.
All methods on the LoadBalancer
interface are called from a Synchronization Context,
meaning they are serialized, thus the balancer implementation doesn't need to worry about
synchronization among them. LoadBalancer.Helper.getSynchronizationContext()
allows implementations to schedule tasks to be run in the same Synchronization Context, with or
without a delay, thus those tasks don't need to worry about synchronizing with the balancer
methods.
However, the actual running thread may be the network thread, thus the following rules must be followed to prevent blocking or even dead-locking in a network:
handleSubchannelState()
) while calling into another class that may involve locks, be cautious
of deadlock. Generally you wouldn't need any locking in the LoadBalancer.A LoadBalancer
keeps states like the latest addresses from NameResolver, the
Subchannel(s) and their latest connectivity states. These states are mutated within the Channel
Executor.
A typical SubchannelPicker
holds a snapshot of these states. It may
have its own states, e.g., a picker from a round-robin load-balancer may keep a pointer to the
next Subchannel, which are typically mutated by multiple threads. The picker should only mutate
its own state, and should not mutate or re-acquire the states of the LoadBalancer. This way the
picker only needs to synchronize its own states, which is typically trivial to implement.
When the LoadBalancer states changes, e.g., Subchannels has become or stopped being READY, and
we want subsequent RPCs to use the latest list of READY Subchannels, LoadBalancer would create
a new picker, which holds a snapshot of the latest Subchannel list. Refer to the javadoc of
handleSubchannelState()
how to do this properly.
No synchronization should be necessary between LoadBalancer and its pickers if you follow the pattern above. It may be possible to implement in a different way, but that would usually result in more complicated threading.
Modifier and Type | Class and Description |
---|---|
static class |
LoadBalancer.Factory
Factory to create
LoadBalancer instance. |
static class |
LoadBalancer.Helper
Provides essentials for LoadBalancer implementations.
|
static class |
LoadBalancer.PickResult
A balancing decision made by
SubchannelPicker for an RPC. |
static class |
LoadBalancer.PickSubchannelArgs
Provides arguments for a
LoadBalancer.SubchannelPicker.pickSubchannel(
LoadBalancer.PickSubchannelArgs) . |
static class |
LoadBalancer.Subchannel
A logical connection to a server, or a group of equivalent servers represented by an
EquivalentAddressGroup . |
static class |
LoadBalancer.SubchannelPicker
The main balancing logic.
|
Constructor and Description |
---|
LoadBalancer() |
Modifier and Type | Method and Description |
---|---|
abstract void |
handleNameResolutionError(Status error)
Handles an error from the name resolution system.
|
abstract void |
handleResolvedAddressGroups(List<EquivalentAddressGroup> servers,
Attributes attributes)
Handles newly resolved server groups and metadata attributes from name resolution system.
|
abstract void |
handleSubchannelState(LoadBalancer.Subchannel subchannel,
ConnectivityStateInfo stateInfo)
Handles a state change on a Subchannel.
|
abstract void |
shutdown()
The channel asks the load-balancer to shutdown.
|
public abstract void handleResolvedAddressGroups(List<EquivalentAddressGroup> servers, @NameResolver.ResolutionResultAttr Attributes attributes)
servers
contained in EquivalentAddressGroup
should be considered equivalent
but may be flattened into a single list if needed.
Implementations should not modify the given servers
.
servers
- the resolved server addresses, never empty.attributes
- extra information from naming system.public abstract void handleNameResolutionError(Status error)
error
- a non-OK statuspublic abstract void handleSubchannelState(LoadBalancer.Subchannel subchannel, ConnectivityStateInfo stateInfo)
The initial state of a Subchannel is IDLE. You won't get a notification for the initial IDLE state.
If the new state is not SHUTDOWN, this method should create a new picker and call Helper.updateBalancingState()
. Failing to do so may result in
unnecessary delays of RPCs. Please refer to PickResult.withSubchannel()
's javadoc for more information.
SHUTDOWN can only happen in two cases. One is that LoadBalancer called LoadBalancer.Subchannel.shutdown()
earlier, thus it should have already discarded this Subchannel. The other
is that Channel is doing a forced shutdown
or has already
terminated, thus there won't be further requests to LoadBalancer. Therefore, SHUTDOWN can be
safely ignored.
subchannel
- the involved SubchannelstateInfo
- the new statepublic abstract void shutdown()