Github user bowenli86 commented on a diff in the pull request:
--- Diff: docs/ops/state/large_state_tuning.md ---
@@ -234,4 +234,97 @@ Compression can be activated through the
**Notice:** The compression option has no impact on incremental snapshots,
because they are using RocksDB's internal
format which is always using snappy compression out of the box.
+## Task-Local Recovery
+In Flink's checkpointing, each task produces a snapshot of its state that
is then written to a distributed store. Each task acknowledges
+a successful write of the state to the job manager by sending a handle
that describes the location of the state in the distributed store.
+The job manager, in turn, collects the handles from all tasks and bundles
them into a checkpoint object.
+In case of recovery, the job manager opens the latest checkpoint object
and sends the handles back to the corresponding tasks, which can
+then restore their state from the distributed storage. Using a distributed
storage to store state has two important advantages. First, the storage
+is fault tolerant and second, all state in the distributed store is
accessible to all nodes and can be easily redistributed (e.g. for rescaling).
+However, using a remote distributed store has also one big disadvantage:
all tasks must read their state from a remote location, over the network.
+In many scenarios, recovery could reschedule failed tasks to the same task
manager as in the previous run (of course there are exceptions like machine
+failures), but we still have to read remote state. This can result in
*long recovery times for large states*, even if there was only a small failure
+a single machine.
+Task-local state recovery targets exactly this problem of long recovery
times and the main idea is the following: for every checkpoint, we do not
+only write task states to the distributed storage, but also keep *a
secondary copy of the state snapshot in a storage that is local to the task*
+(e.g. on local disk or in memory). Notice that the primary store for
snapshots must still be the distributed store, because local storage does not
+ensure durability under node failures abd also does not provide access for
other nodes to redistribute state, this functionality still requires the
+However, for each task that can be rescheduled to the previous location
for recovery, we can restore state from the secondary, local
+copy and avoid the costs of reading the state remotely. Given that *many
failures are not node failures and node failures typically only affect one
+or very few nodes at a time*, it is very likely that in a recovery most
tasks can return to their previous location and find their local state intact.
+This is what makes local recovery effective in reducing recovery time.
+Please note that this can come at some additional costs per checkpoint for
creating and storing the secondary local state copy, depending on the
+chosen state backend and checkpointing strategy. For example, in most
cases the implementation will simply duplicate the writes to the distributed
+store to a local file.
+<img src="../../fig/local_recovery.png" class="center" width="80%"
alt="Illustration of checkpointing with task-local recovery."/>
+### Relationship of primary (distributed store) and secondary (task-local)
+Task-local state is always considered a secondary copy, the ground truth
of the checkpoint state is the primary copy in the distributed store. This
+has implications for problems with local state during checkpointing and
+- For checkpointing, the *primary copy must be successful* and a failure
to produce the *secondary, local copy will not fail* the checkpoint. A
+will fail if the primary copy could not be created, even if the secondary
copy was successfully created.
+- Only the primary copy is acknowledged and managed by the job manager,
secondary copies are owned by task managers and their life cycle can be
+independent from their primary copy. For example, it is possible to retain
a history of the 3 latest checkpoints as primary copies and only keep
+the task-local state of the latest checkpoint.
+- For recovery, Flink will always *attempt to restore from task-local
state first*, if a matching secondary copy is available. If any problem occurs
+the recovery from the secondary copy, Flink will *transparently retry to
recovery the task from the primary copy*. Recovery only fails, if primary
+and the (optional) secondary copy failed. In this case, depending on the
configuration Flink could still fall back to an older checkpoint.
+- It is possible that the task-local copy contains only parts of the full
task state (e.g. exception while writing one local file). In this case,
+Flink will first try to recover local parts locally, non-local state is
restored from the primary copy. Primary state must always be complete and is
+a *superset of the task-local state*.
+- Task-local state can have a different format than the primary state,
they are not required to be byte identical. For example, it could even possible
+the task-local state is an in-memory consisting of heap objects, and not
stored in any files.
+- If a task manager is lost, the local state from all its task is lost.
+### Configuring task-local recovery
+Task-local recovery is *deactivated by default* and can be activated
through Flink's configuration with the key `state.backend.local-recovery` as
+in `CheckpointingOptions.LOCAL_RECOVERY`. Users have currently two choices:
+- `DISABLED`: Local recovery is disabled (default).
+- `ENABLE_FILE_BASED`: Local recovery is activated, based on writing a
secondary copy of the task state on local disk.
+### Details on task-local recovery for different state backends
+***Limitation**: Currently, task-local recovery only covers keyed state
backends. Keyed state is typically by far the largest part of the state. In the
near future, we will
+also cover operator state and timers.*
+The following state backends can support task-local recovery.
+- FsStateBackend: task-local recovery is supported for keyed state. The
implementation will duplicate the state to a local file. This can introduce
additional write costs
+and occupies local disk space. In the future, we might also offer an
implementation that keeps task-local state in memory.
--- End diff --