You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 18 Next »

Choosing algorithm for transactional consistency

There are some possible solutions to guarantee transactional consistency:

  1. Consistent Cut
    + Simple algorithm - requires only some read-write locks to sync threads, and doesn't affect performance much.
    + Doesn't require to much additional space - it just writing few additional messages to WAL.
    - Requires time to recovery (applying every message from WAL to system) that depends on how many operations need to be restored.
    - Requires additional storage to persist WAL archives.
    - Doesn't restore ATOMIC caches.
  2. Incremental physical snapshots (collection of partition binary files changed since previous snapshot, can be implemented as delta or as full copy).
    - High disk IO usage for preparing snapshots.
    - Current implementation of snapshots requires PME, that affects performance much.
    + Fast recovering, doesn't depend on amount operations to restore (WAL-free).
  3. MVCC
    - Ignite failed to support MVCC, very hard to implement.

From those options Consistent Cut is more preferable. It may require optimizations for recovery, but it looks like we can provide some. For example, some options:

  1. Using WAL compaction for archived files, that excludes physical records from WAL files;
  2. Apply DataEntry from WAL in parallel by using striped executor (cache group id and partition id);
  3. Using index over WAL files for fast access to written Consistent Cuts.

Consistent Cut

Goal is to find Point on timeline on every node such that set of committed transactions before (after) this Point is the same on every node in a cluster.

Consider a simple model. On the pictures below there are two lines (P and Q). They represent two different Ignite nodes. Nodes exchange of transaction messages with each other. For simplicity let's consider there is the only state of tx (COMMITTED) and only single message related to one tx (one node notifies another about tx and latter immediately commits it after receiving the message).

Transaction can be described as ordered sequence of events. Correct order of events is known on every node. There are 4 ordered events to describe transaction in this simple model: 

  1. P.tx(COMMITTED): transaction committed on node P;
  2. P.snd(m): send message msg(tx5: P → Q);
  3. Q.rcv(m): receive message msg(tx5: P → Q);
  4. Q.tx(COMMITTED): transaction committed on node Q.

The cut line crosses both lines (P and Q) in the Points. Transactions (tx1, tx2, tx3, tx4) are on the left side of cut on both nodes. But tx5 is different because cut crosses also a message line. Then tx5 commits on different nodes in different states (before / after cut). To make a cut consistent the global order of transaction events must be kept - for any transaction if AFTER state contains some events from this order it means that BEFORE must contain all prior events.

Consider examples. First picture below describes case where tx5 state is:

  • before cut: node P = { }, node Q = { rcv(m), tx(COMMITTED) }.
  • after cut: node P = { tx(COMMITTED), snd(m) }, node Q = { }.

Global state before cut contains events { Q.rcv(m), Q.tx(COMMITTED) } that aren't leading events in the known sequence - it misses first events { P.tx(COMMITTED), P.snd(m) }. But order of events must match order of consistent states (BEFORE, AFTER), then such a cut is inconsistent.  

Second picture below describes case where tx5 state is:

  • before cut: node P = { tx(COMMITTED), snd(m) }, node Q = { }.
  • after cut: node P = { }, node Q = { rcv(m), tx(COMMITTED) }.

Global state before cut contains events { P.tx(COMMITTED), P.snd(m) } that are leading events in the known sequence. Such a cut is consistent.


In case of consistent cut every node knows which side of cut a local transaction belongs to (BEFORE or AFTER). On the second picture, node Q after event rcv knows whether tx5 was included to cut on node P. Then if it was, node Q includes it, otherwise doesn't include. 

Algorithm

There is a paper [1] that proposes an algorithm for implementing distributed consistent cut. This algorithm works without a coordinator of the procedure, it is suitable for processes with no-FIFO channels between processes.

Algorithm's purpose is identify wrong order of messages by signing the messages with actual color (red or white). Then every node knows which side of the cut this message belongs to. 

Messages:

  1. sent - a set of messages that changed state of a sender.
  2. received - a set of messages that changed state of receiver.

Algorithm's steps:

  1. Initially all process are white, sent and received collections are empty, LocalState is empty.
  2. After some time of system work, every node might have:
    1. Optionally empty collections sent and received
    2. Optionally non-empty LocalState <-> sent + received. State match events that changed its state.
  3. Random process can start a snapshot (furthermore, multiple process may start it simultaneously):
    1. Node colors itself to red.
    2. It commits a LocalState.
    3. It commits sent and received as collections for every IN and OUT channel. New one created for next LocalState.
    4. It prepares a marker message: it is red, and has a payload of sent. Goal of the marker is to guarantee order of messages (receivedij must be a subset of sentji).
  4. Mark every ordinal message between distributed processes with the marker message, if no upcoming message to a node, then it just sends the marker as an ordinary message.
  5. On receiving the ordinal message a process has to check the marker at first, before applying the message;
  6. If receiving color differs from local color, node has to trigger the local snapshot procedure.
  7. Handle sent from the received marker:
    1. calculates ChannelState for the channel it received a message: sent - received; where sent extracts from the marker, received - calculates locally since local snapshot.
  8. On received marker messages from all IN channels, it prepares a snapshot:
    1. Local snapshot of node i: Ni = LocalStatei + Σ ChannelStateij (sent - received)
  9. Every such local snapshot is a unit of global snapshot:
    1. Note, that snapshot consist of committed LocalStates and messages between nodes.
    2. committed sent and received collections are cleaned.

Map the algorithm to Ignite

Ignite components

On receiving a message with new CutVersion node sets it and commits LocalState and ChannelState - to identify wrong order of the events. 

  1. LocalState maps to local WAL (all of committed transactions are part of LocalState);
  2. Channel:
    1. We can piggy back on Ignite transaction protocol (Prepare, Finish) messages with CommunicationSpi.
    2. In case there is no transaction for a channel, we can rely on the DiscoverySpi to start local snapshot on non-participated nodes.
  3. ChannelState maps to `IgniteTxManager#activeTransactions`:
    1. sent collection match committed transactions for which local node is near - they send FinishMessages to other nodes.
    2. received collection match committed transactions for which local node isn't near - they receive FinishMessages from other nodes. 
  4. `IgniteTxManager#activeTransactions` doesn't track:
    1. committing transactions (COMMITTING+), they are removed from this collection before start committing them.
      1. track them additionally: add to a separate collection before it starts committing, and remove after it committed.

Naive algorithm

  1. Initial state:
    1. Ignite nodes started from snapshot, or other consistent state (after graceful cluster stop / deactivation).
    2. var color = WHITE .
    3. two empty collections for sent and received - Map<UUID, Collection<GridCacheVersion>>, where key is a remote node ID, value is nearXidVersion of committed transaction.
  2. After some time, Ignite nodes might have non-empty send and received collections.
  3. Ignite coordinator starts a local snapshot (it all must be done atomically):
    1. mark local var with color = RED 
    2. write a snapshot record to WAL (commits LocalState).
    3. replace sent and received for new empty instances, old collections are immutable now for the snapshot process.
    4. prepare a marker message to every node with subset of send transactions (send just Collection<GridCacheVersion> for every node).
  4. Send marker messages directly to all nodes in a cluster. It must be sent with every next FinishMessage (to avoid cases, when next FinishMessage received before marker message).
  5. Other nodes on receiving a marker with new color:
    1. starts a local snapshot (see steps from point 3).
  6. For every received marker message:
    1. For prepared received collection and received marker's sent calculate: (sent - received). It transactions that are in transit state. They are also part of snapshot.
  7. After markers from all nodes received:
    1. Will store ChannelState into WAL: write Map<UUID, Collection<GridCacheVersion>>, where value is (sent - received).
    2. Snapshot is finished locally.
  8. For all nodes color = RED . Next snapshot iteration will started with changing color to WHITE . 

Disadvantages:

  1. Collections sent and received could be too big to track and send with single message.
  2. It requires a lock for steps 3 and 4, it affects of sending FinishMessage that affects UX (lower performance of transactions).
  3. It worst case it requires O(N^2) additional messages, when N is cluster size. It's possible when no active transactions are in cluster (idle nodes).

Reducing the sent and received collections:

We can rely on two-phase protocol of Ignite transactions to reduce sent and received:

  1. Transactions have additional state (PREPARED) before sending and receiving a FinishMessage. Then we can say that PREPARED is a promise for sending/receiving FinishMessage.
  2. Then IgniteTxManager#activeTransactions  already can replace sent and received collections:
    1. it tracks all PREPARED transactions (it is analogue of sent - received).
    2. Also, it automatically shrinks after transaction committed.

Avoid locking in steps 3 and 4

  1. Removing writing WAL record from lock => inconsistency between fixed sent / received and committed LocalState. 
  2. Removing fixing sent and received from lock => inconsistency between fixed sent / received and committed LocalState.

To avoid misses:

  • sent and received might be wider - collect transaction that not only PREPARED, but PREPARING+. Do not track ACTIVE transactions (as algorithm guarantees they will be excluded from snapshot: ACTIVE < PREPARED than such txs aren't part neither of ChannelState or LocalState).
  • auto-shrink must be disabled during a snapshot process.

To avoid unwanted transactions:

  • Introduce a new collection excluded that is a collection of messages that changed LocalState concurrently with local snapshot, but they aren't part of the snapshot ChannelState (but part of next ChannelState, after snapshot).
  • Then Ignite need a rule to distinct different ChannelStates (before and after snapshot) - add additional flag to FinishMessage that shows which side of snapshot this transaction belongs to on near node.

Also there is no need to send send collection with FinishMessage. Then whole lock is reduced to single volatile variable:

  1. Atomically (with Compare-And-Swap) update the color variable and disable auto-shrink for active transactions.
  2. Write a LocalState to WAL
  3. Collect all active transactions
  4. FinishMessage is sent with 2 fields: color (snp state) and commit color for the specified transaction.
  5. Introduce the exclude collection that contains of transactions to exclude from Snapshot before LocalState commit while restoring.

Reducing additional messages

As now useful info is just a color field (no need to send sent collection) then:

  1. for idle nodes it's cheaper to send a single message over discovery, complexity O(N), where N - cluster size.
  2. no need to await all marker messages from other nodes. Every node can just dump active transactions.

Final algorithm

For Ignite implementation it's proposed to use only single node to coordinate algorithm. User starts a command for creating new ConsistentCut on single node and this node becomes responsible for coordinating Consistent Cut iteration:

  1. Initial state:
    1. Ignite nodes started from snapshot, or other consistent state (after graceful cluster stop / deactivation).
    2. Every Ignite node holds a color (RED / WHITE / null), initially null.
    3. Empty collection committingTxs (Set<GridCacheVersion>) that goal is to track COMMITTING+ transactions, that aren't part of IgniteTxManager#activeTx . It's automatically shrinks after transaction committed.
  2. After some time, Ignite nodes might have non-empty committingTxs.
  3. Ignite node inites a global snapshot, by starting DistributedProcess (by discovery IO):
    1. switches a color (null → WHITE → RED → WHITE → ...)
    2. creates a new timestamp, checks that it's greater than previous (incremental snapshot).
    3. prepares a marker message that contains the color  and the timestamp. And transmits this message to other nodes.
  4. Every nodes starts a local snapshot process after receiving the marker message (whether by discovery, or by communication with transaction message) 
    1. Atomically: updates local `color`, disable auto-shrink of committingTxs, prepare ConsistentCut future.
    2. Write a snapshot record to WAL with the received timestamp (commits LocalState).
    3. Collect of active transactions - concat of IgniteTxManager#activeTx and committingTxs 
    4. Prepares 2 empty collections - before [sent - received] and after [exclude] cut.
  5. While global Consistent Cut is running every node signs output transaction messages:
    1. The marker message.
    2. Finish messages is signed on node that commits first (near node for 2PC, backup or primary for 1PC) with color  to notify other nodes which side of cut transaction belongs to.
  6. For every collected active transaction, node waits for Finish message, to extract the color and fills before, after collections:
    1. if received color is null or differs from local, then transaction on before side
    2. if received color equals to local, then transaction on after side
  7. After all transactions finished:
    1. Writes a WAL record with ChannelState (before, after). 
    2. Clears committingTxs and enables auto-shrink again.
    3. Completes ConsistentCut future, and notifies a node-initiator about finishing local procedure (with DistributedProcess protocol).
  8. After all nodes finished ConsistentCut, every node stops signing outgoing transaction messages.
  9. Every node, now have updated color (non-null).

Consistent and inconsistent Cuts

Consistent Cut, in terms of Ignite implementation, 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:

  1. any errors appeared during processing local Cut.
  2. if a transaction is recovered with transaction recovery protocol (tx.finalizationStatus == RECOVERY_FINISH).
  3. if transaction finished in UNKNOWN state.
  4. baseline topology change, Ignite nodes finishes local Cuts running in this moment, making them inconsistent.

ConsistentCutVersion

Every ignite nodes tracks current ConsistentCutVersion:

ConsistentCutVersion
class ConsistentCutVersion {
	long version;
}

`version` is a simple counter. It's guaranteed it is raising monotonically, due to it is incremented by discovery communication.

ConsistentCutColor initialization

ConsistentCutColor can be initialized with:

For changed server topology: 

  1. The color is received with GridDescoveryData from coordinator (for both client and server nodes).
  2. Every server node on receiving TcpDiscoveryNodeAddedMessage:
    1. finishes local ConsistentCut.

Order of transaction messages

In the 2PC protocol in most cases (except some optimization and corner cases) sequence of events for transaction can be described as set of messages (PrepareRequest, PrepareResponse, FinishRequest, FinishResponse) and set of TransactionState. For case with 2 nodes (P - near, Q - primary) it looks like that:  

P.tx(PREPARING) → P.snd(GridNearPrepareRequest) → Q.rcv(GridNearPrepareRequest) → Q.tx(PREPARED) → Q.snd(GridNearPrepareResponse) → P.rcv(GridNearPrepareResponse) → P.tx(PREPARED) → P.tx(COMMITTED) → P.snd(GridNearFinishRequest) →  Q.rcv(GridNearFinishRequest) → Q.tx(COMMITTED)


Important steps in this sequence:

  • near node commits before primary node (primary node commits before backup node)
  • there is a FinishRequest between between actual commits.

Consistent Cut procedure can start in any moment and cross this sequence in any chain. In this moment every node should decide which side of ConsistentCut every transaction belongs to. Some important notes:

  1. There are some issues with ACTIVE transactions:
    1. They can be long
    2. They may become SUSPENDED and hang ConsistentCut after that
  2. Then it's better to avoid checking such transactions, and algorithm helps it: for every transaction in ACTIVE state there are at least 2 messages to sync CutVersion between nodes (PREPARE response, FINISH request) - it's enough to provide a guarantee that every ACTIVE transaction will on the AFTER side.
  3. We skip SUSPENDED transactions to avoid a Consistent Cut hangs. The only valid way to reach SUSPENDED state is ACTIVE → SUSPENDED. Then no need additional checks there.
  4. For other transactions (>= PREPARING) algorithm listens their finish futures and checks txCutVer. 

Consider some examples below (P - near, Q - primary):

  1. First picture: P commits tx BEFORE, Q commits tx AFTER. Globally tx BEFORE cut.
    1. P commits tx BEFORE, sign FinishRequest with latest known CutVersion (before new one).
    2. Q collects this tx in moment of cut (as locally it's PREPARED) and waiting for FinishRequest. On receiving it checks message: local CutVer > rcvd txCutVer, then it belongs to BEFORE.

  2. Second picture: P commits tx AFTER, Q commits tx AFTER. Globally tx AFTER cut.
    1. P: cut before rcv(GridNearPrepareResponse), then tx isn't PREPARED yet. Then P signs FinishRequest with new CutVersion.
    2. Q collects this tx in moment of cut (as locally it's PREPARED) and waiting for FinishRequest. On receiving it checks message: local CutVer == rcvd txCutVer, then belongs AFTER.

Pictures below describes a little bit complicate example: P commits tx AFTER, Q commits tx BEFORE. But actually Consistent Cut algorithm doesn't allow such case, because this cut is inconsistent. Global state of tx includes events:  { Q.snd(GridNearPrepareResponse), Q.rcv(GridNearFinishRequest), Q.tx(COMMITTED) }. But this state misses some middle events { P.rcv(GridNearPrepareResponse), P.tx(COMMITTED), P.snd(GridNearFinishRequest) }.

Then this cut is inconsistent. To avoid such cases, FinishRequest is signed with latestCutVer, and node Q must to validate it and trigger Consistent Cut before applying the message. And after that this case equals to case from previous pictures.

  1. P: cut before P.rcv(GridNearPrepareResponse), then tx isn't PREPARED yet. Then P signs FinishRequest with new CutVersion.
  2. Q: by receiving FinishRequest it triggers Consistent Cut, collect PREPARED tx in this moment, and after that check version from FinishRequest.
  3. Case equals to that on the second picture above. Globally tx AFTER cut.

One-Phase commit handling

Sequence of events for 1PC differs:

P.tx(PREPARING) → P.snd(GridNearPrepareRequest) → P.rcv(GridNearPrepareRequest) → Q.tx(PREPARED) → Q.tx(COMMITTED) → Q.snd(GridNearPrepareResponse) → P.rcv(GridNearPrepareResponse) → P.tx(PREPARED) → P.tx(COMMITTED)

Important points in this sequence

  • backup node commits before primary node
  • there is a GridNearPrepareResponse between between actual commits 

Then for 1PC the backup (or primary, if backups=0) is responsible for signing tx with CutVersion, not near node. CutVersion propogates between nodes in reverse order: near ← primary ← backup.

Signing messages

Ignite transaction protocol includes multiple messages. But only some of them affects meaningful (relating to the algorithm) that change state of transactions (PREPARED, COMMITTED):

  1. GridNearTxPrepareRequest / GridDhtTxPrepareRequest
  2. GridNearTxPrepareResponse / GridDhtTxPrepareResponse
  3. GridNearTxFinishRequest / GridDhtTxFinishRequest

Also some messages require to be signed with tx color to check it them on primary/backup node:

  1. GridNearTxFinishRequest / GridDhtTxFinishRequest
  2. GridNearTxPrepareResponse / GridDhtTxPrepareResponse (for 1PC algorithm).

Unstable topology

There are some cases to handle for unstable topology:

  1. Client or non-baseline server node leaves – no need to handle.
  2. Server node leaves:
    1. all nodes finish local running ConsistentCut, making them in-consistent
  3. Server node joins:
    1. all nodes finish local running ConsistentCut, making them in-consistent
    2. new node checks whether rebalance was required for recovering. If it is required, then handle it TBD

TBD: Which ways to use to avoid inconsistency between data and WAL after rebalance. There are options:

  1. First solution: set flag ConsistentCutManager#inconsistent  to true , and persist this flag within local MetaStorage.
    1. On receiving new ConsistentCutVersion check the flag and raise an exception.
    2. Clean flag on recovering from ClusterSnapshot, or after creating a ClusterSnapshot.
    3. + Simple handling in runtime.
    4. - Need to rebalance a single node after PITR with file-based (or other?) rebalance: more time for recovery, file-based rebalance has not-resolved issues yet(?). 

  2. Do not disable WAL during rebalance:
    + WAL now is consistent with data, can use it for PITR after rebalancing
    - Too many inserts during rebalance may affect rebalance speed, IO utilization

  3. Automatically create snapshots for rebalancing cache groups:
    + Guarantee of consistency (snapshot is created on PME, after rebalance).
    + Faster recovery.
    - Complexity of handling - Ignite should provide additional tool for merging WAL and multiple snapshots created in different time
    ? is it possible to create snapshot only on single rebalanced node?
    ? Is it possible to sync current PME (on node join) with starting snapshot?

  4. Write a special Rebalance record to WAL with description of demanded partition:
    1. During restore read this record and repeat the historical rebalance at this point, after rebalance resume recovery with existing WALs.
    2. In case Record contains full rebalance - stops recovering with WAL and fallback to full rebalance.
      ? Is it possible to rebalance only specific cache groups, and continue to WAL recovery for others.
      - For historical rebalance during recovery need separate logic for extracting records from WAL archives from other nodes. 

WAL records

There are 2 records: `ConsistentCutStartRecord` for Start event and `ConsistentCutFinishRecord` for Finish event. 

  • ConsistentCutStartRecord: record is written to WAL in moment when CC starts on a local node. It helps to limit amout of active transactions to check. But there is no strict guarantee for all transactions belonged to the BEFORE side to be physically committed before ConsistentCutStartRecord, and vice versa. This is the reason for having ConsistentCutFinishRecord.
  • ConsistentCutFinishRecord: This record is written to WAL after Consistent Cut stopped analyzing transactions and storing them in a particular bucket (BEFORE or AFTER).

It guarantees that the BEFORE side consist of:
1. transactions physically committed before ConsistentCutStartRecord and weren't included into ConsistentCutFinishRecord#after();
2. transactions physically committed between ConsistentCutStartRecord and ConsistentCutFinishRecord and were included into ConsistentCutFinishRecord#before().

It guarantees that the AFTER side consist of:
1. transactions physically committed before ConsistentCutStartRecord and were included into ConsistentCutFinishRecord#after();
2. transactions physically committed after ConsistentCutStartRecord and weren't included into ConsistentCutFinishRecord#before().


ConsistentCutRecord
/** */
public class ConsistentCutStartRecord extends WALRecord {
	/** Timestamp of starting global Consistent Cut. */
	private final long ts;
}


/** */
public class ConsistentCutFinishRecord extends WALRecord {
    /**
     * Collections of TXs committed BEFORE the ConsistentCut (sent - received).
     */
    private final Set<GridCacheVersion> before;

     /**
     * Collections of TXs committed AFTER the ConsistentCut (exclude).
     */
    private final Set<GridCacheVersion> after;
 }
  1. ON DISTRIBUTED SNAPSHOTS, Ten H. LAI and Tao H. YANG, 29 May 1987
  • No labels

Apache Ignite