libmuscle.manager.snapshot_registry module

class libmuscle.manager.snapshot_registry.SnapshotNode(num: int, instance: ~ymmsl.identity.Reference, snapshot: ~libmuscle.snapshot.SnapshotMetadata, peers: ~typing.FrozenSet[~ymmsl.identity.Reference], consistent_peers: ~typing.Dict[~ymmsl.identity.Reference, ~typing.List[~libmuscle.manager.snapshot_registry.SnapshotNode]] = <factory>)

Bases: object

Represents a node in the snapshot graph.

num

The number of the snapshot. Unique for this instance. Later snapshots always have a higher num.

Type:

int

instance

Which instance this is a snapshot of.

Type:

ymmsl.identity.Reference

snapshot

The snapshot metadata reported by the instance.

Type:

libmuscle.snapshot.SnapshotMetadata

peers

The set of peers that the instance is connected to.

Type:

FrozenSet[ymmsl.identity.Reference]

consistent_peers

Keeps track of snapshots per peer that are consistent with this one.

Type:

Dict[ymmsl.identity.Reference, List[libmuscle.manager.snapshot_registry.SnapshotNode]]

property consistent: bool

Returns True iff there is a consistent checkpoint with all peers.

consistent_peers: Dict[Reference, List[SnapshotNode]]

do_consistency_check(peer_node: SnapshotNode, connections: List[Tuple[Identifier, Identifier, _ConnectionInfo]]) → bool

Check if the snapshot of the peer is consistent with us.

When the peer snapshot is consistent, adds it to our list of consistent peer snapshots (in consistent_peers) and vice versa.

Parameters:

• peer_node – Snapshot of one of our peers

• connections – All connections from our instance to the peer instance

Returns:

True iff the peer snapshot is consistent with ours.

instance: Reference

num: int

peers: FrozenSet[Reference]

snapshot: SnapshotMetadata

class libmuscle.manager.snapshot_registry.SnapshotRegistry(config: PartialConfiguration, snapshot_folder: Path, topology_store: TopologyStore)

Bases: Thread

Registry of all snapshots taken by instances.

Current snapshots are stored in a graph. Every node represents a snapshot taken by an instance (see SnapshotNode). When snapshots from peer instances are consistent, the nodes are connected to each other.

This class manages the snapshot nodes. New snapshots are registered through register_snapshot().

register_snapshot(instance: Reference, snapshot: SnapshotMetadata) → None

Register a new snapshot.

Parameters:

• instance – The instance that created the snapshot

• snapshot – Metadata describing the snapshot

run() → None

Code executed in a separate thread

shutdown() → None

Stop the snapshot registry thread

libmuscle.manager.snapshot_registry.calc_consistency(num1: int, num2: int, first_is_sent: bool, num2_is_restart: bool) → bool

Calculate consistency of message counts.

Parameters:

• num1 – message count of instance 1

• num2 – message count of instance 2

• first_is_sent – True iff instance 1 is sending messages over this conduit

• num2_is_restart – True iff the snapshot of num2 is a full restart

Returns:

True iff the two message counts are consistent

libmuscle.manager.snapshot_registry.calc_consistency_list(num1: List[int], num2: List[int], first_is_sent: bool, num2_is_restart: bool) → bool

Calculate consistency of message counts.

Parameters:

• num1 – message count of instance 1

• num2 – message count of instance 2

• first_is_sent – True iff instance 1 is sending messages over this conduit

• num2_is_restart – True iff the snapshot of num2 is a full restart

Returns:

True iff the two message counts are consistent

libmuscle.manager.snapshot_registry.safe_get(lst: List[_T], index: int, default: _T) → _T

Get an item from the list, returning default when it does not exist.

Parameters:

• lst – List to get the item from

• index – Which item to get, should be >= 0

• default – Value to return when hitting an IndexError