ConsistentCut splits WAL on 2 global areas - Before
and After
. It guarantees that every transaction committed Before
also will be committed Before
on every other node participated in the transaction. It means that an Ignite nodes can safely recover themself to the consistent Before
state without any coordination with each other.
The border between Before
and After
areas consists of two WAL records - ConsistentCutStartRecord
and ConsistentCutFinishRecord
. It guarantees that the Before
consists of:
ConsistentCutStartRecord
and weren't included into ConsistentCutFinishRecord#after().
ConsistentCutStartRecord
and ConsistentCutFinishRecord
and were included into ConsistentCutFinishRecord#before()
.On the picture below the Before
area consist of transactions colored to yellow, while After
is green.
/** */ public class ConsistentCutStartRecord extends WALRecord { /** Consistent Cut ID. */ private final UUID cutId; } /** */ public class ConsistentCutFinishRecord extends WALRecord { /** Consistent Cut ID. */ private final UUID cutId; /** * Collections of transactions committed BEFORE. */ private final Set<GridCacheVersion> before; /** * Collections of transactions committed AFTER. */ private final Set<GridCacheVersion> after; }
lastFinishedCutId
holds previous ConsistentCutId
, or null.DistributedProcess
with special message holds new ConsistentCutId
(goal is to notify every node in a cluster about running incremental snapshot). topVer
on which ConsistentCut started.ConsistentCutId
by discovery.ConsistentCutId
by communication (transaction messages - Prepare, Finish).ConsistentCutId
, every node: topVer
with mapped topology version of received transaction message. Skip ConsistentCut initialization if the received version is higher.ConsistentCut
future.committingTx
(goal is to track COMMITTING transactions, that aren't part of IgniteTxManager#activeTx
)ConsistentCutId
.IgniteTxManager#activeTx.
Set listeners on those tx#finishFuture.
ConsistentCutStartRecord
to WAL with the received ConsistentCutId
.committingTxs.
Set listeners on those tx#finishFuture.
committingTxs
to null.
DistributedProcess
is running every node is signing output transaction messages:ConsistentCutId
(to trigger ConsistentCut
on remote node, if not yet).txCutId
on the node that commits first (if it's not null then transaction starts committing After
Consistent Cut):committingTxs
, and marks the transaction with extracted from the message txCutId.
check If transaction state is UNKNOWN or status is RECOVERY_FINISH, then complete ConsistentCut with exception.
topVer
, then put it into after
.txCutId
equals to local, then put transaction into after
, otherwise put into before
.ConsistentCutFinishRecord
into WAL with the collections ( before
, after
). ConsistentCut
future.ConsistentCut,
on every node:lastFinishedCutId
with the current id.
ConsistentCut
future becomes null.Consistent Cut is such cut that correctly finished on all baseline nodes - ConsistentCutStartRecord
and ConsistentCutFinishRecord
are written.
"Inconsistent" Cut is such a cut when one or more baseline nodes hasn't wrote ConsistentCutFinishRecord
. It's possible in cases:
tx.finalizationStatus
== RECOVERY_FINISH).On the picture below on left side is a diagram of sending transaction messages. Before sending message it checks whether cut is running. If it is then wraps the message, otherwise send ordinary message (PrepareRequest
in example).
Ignite transaction protocol includes multiple messages. But only some of them affects meaningful (relating to the algorithm) that change state of transactions (PREPARED, COMMITTED):
GridNearTxPrepareRequest / GridDhtTxPrepareRequest
GridNearTxPrepareResponse / GridDhtTxPrepareResponse
GridNearTxFinishRequest / GridDhtTxFinishRequest
Those messages are wrapped in ConsistentCutAwareMessage
that is prepared right before sending message on other node. They used the current ConsistentCutId
.
class ConsistentCutAwareMessage { /** Original transaction message. */ Message msg; /** Consistent Cut ID. */ UUID cutId; /** Cluster topology version on which Consistent Cut started. */ long topVer; }
Also some messages require to be signed with additional ConsistentCutId
to check it them on primary/backup node.
GridNearTxFinishRequest / GridDhtTxFinishRequest
GridNearTxPrepareResponse / GridDhtTxPrepareResponse
(for 1PC algorithm).Those messages are wrapped in ConsistentCutAwareTxFinishMessage
that is prepared right before transaction starts committing on first committing node. They used the current ConsistentCutId
for this setting. If current ConsistentCutId
is not null, then transaction starts committing after ConsistentCut started and it means that this transaction belongs the After
side.
class ConsistentCutAwareTransactionFinishMessage extends ConsistentCutAwareMessage { /** Consistent Cut ID after which transaction committed. */ @Nullable UUID txCutId; }