Transaction T accessed data at resource managers R1, … Rn.
The goal is to either
commit T at all of R1, … Rn, or
abort T at all of R1, … Rn
even if resource managers, nodes and communications links fail during the commit or abort activity
That is, never commit at Ri but abort at Rk.
7.2 Two-Phase Commit
Two phase commit (2PC) is the standard protocol for making commit and abort atomic
Coordinator - the component that coordinates commitment at home(T)
Participant - a resource manager accessed by T
A participant P is ready to commit T if all of T’s after-images at P are in stable storage
The coordinator must not commit T until all participants are ready
If P isn’t ready, T commits, and P fails, then P can’t commit when it recovers.
The Protocol
(Begin Phase 1) The coordinator sends a Request-to-Prepare message to each participant
The coordinator waits for all participants to vote
Each participant
votes Prepared if it’s ready to commit
may vote No for any reason
may delay voting indefinitely
(Begin Phase 2) If coordinator receives Prepared from all participants, it decides to commit. (The transaction is now committed.) Otherwise, it decides to abort.
The Protocol (cont’d)
The coordinator sends its decision to all participants (i.e. Commit or Abort)
Participants acknowledge receipt of Commit or Abort by replying Done.
Case 1: Commit
Performance
In the absence of failures, 2PC requires 3 rounds of messages before the decision is made
Request-to-prepare
Votes
Decision
Done messages are just for bookkeeping
they don’t affect response time
they can be batched
Uncertainty
Before it votes, a participant can abort unilaterally
After a participant votes Prepared and before it receives the coordinator’s decision, it is uncertain. It can’t unilaterally commit or abort during its uncertainty period.
Uncertainty (cont’d)
The coordinator is never uncertain
If a participant fails or is disconnected from the coordinator while it’s uncertain, at recovery it must find out the decision
The Bad News Theorems
Uncertainty periods are unavoidable
Blocking - a participant must await a repair before continuing. Blocking is bad.
Theorem 1 - For every possible commit protocol (not just 2PC), a communications failure can cause a participant to become blocked.
Independent recovery - a recovered participant can decide to commit or abort without communicating with other nodes
Theorem 2 - No commit protocol can guarantee independent recovery of failed participants
7.3 2PC Failure Handling
Failure handling - what to do if the coordinator or a participant times out waiting for a message.
Remember, all failures are detected by timeout
A participant times out waiting for coordinator’s Request-to-prepare.
It decides to abort.
The coordinator times out waiting for a participant’s vote
It decides to abort
2PC Failure Handling (cont’d)
A participant that voted Prepared times out waiting for the coordinator’s decision
It’s blocked.
Use a termination protocol to decide what to do.
Naïve termination protocol - wait till the coordinator recovers
The coordinator times out waiting for Done
it must resolicit them, so it can forget the decision
Forgetting Transactions
After a participant receives the decision, it may forget the transaction
After the coordinator receives Done from all participants, it may forget the transaction
A participant must not reply Done until its commit or abort log record is stable
Else, if it fails, then recovers, then asks the coordinator for a decision, the coordinator may not know
If the coordinator fails and later recovers, it must know the decision. It must therefore log
the fact that it began T’s 2PC protocol, including the list of participants, and
Commit or Abort, before sending Commit or Abort to any participant (so it knows whether to commit or abort after it recovers).
If the coordinator fails and recovers, it resends the decision to participants from whom it doesn’t remember getting Done
If the participant forgot the transaction, it replies Done
The coordinator should therefore log Done after it has received them all.
Participant Recovery
If a participant P fails and later recovers, it first performs centralized recovery (Restart)
For each distributed transaction T that was active at the time of failure
If P is not uncertain about T, then it unilaterally aborts T
If P is uncertain, it runs the termination protocol (which may leave P blocked)
To ensure it can tell whether it’s uncertain, P must log its vote before sending it to the coordinator
To avoid becoming totally blocked due to one blocked transaction, P should reacquire T’s locks during Restart and allow Restart to finish before T is resolved.
Heuristic Commit
Suppose a participant recovers, but the termination protocol leaves T blocked.
Operator can guess whether to commit or abort
Must detect wrong guesses when coordinator recovers
Must run compensations for wrong guesses
Heuristic commit
If T is blocked, the local resource manager (actually, transaction manager) guesses
At coordinator recovery, the transaction managers jointly detect wrong guesses.
7.4 2PC Optimizations and Variations
Optimizations
Read-only transaction
Presumed Abort
Transfer of coordination
Cooperative termination protocol
Variations
Re-infection
Phase Zero
Read-only Transaction
A read-only participant need only respond to phase one. It doesn’t care what the decision is.
It responds Prepared-Read-Only to Request-to-Prepare, to tell the coordinator not to send the decision
Limitation - All other participants must be fully terminated, since the read-only participant will release locks after voting.
No more testing of SQL integrity constraints
No more evaluation of SQL triggers
Presumed Abort
After a coordinator decides Abort and sends Abort to participants, it forgets about T immediately.
Assume coordinator includes a list of participants in Request-to-Prepare.
If a participant times-out waiting for the decision, it runs the following protocol.
1. Participant P sends Decision-Req to other participants
2. If participant Q voted No or hasn’t voted or received Abort from the coordinator, it responds Abort
3. If participant Q received Commit from the coordinator, it responds Commit.
4. If participant Q is uncertain, it responds Uncertain (or doesn’t respond at all).
If all participants are uncertain, then P remains blocked.
Cooperative Termination Issues
Participants don’t know when to forget T, since other participants may require CTP
Solution 1 - After receiving Done from all participants, coordinator sends End to all participants
Solution 2 - After receiving a decision, a participant may forget T any time.
To ensure it can run CTP, a participant should include the list of participants in the vote log record.
Reinfection
Suppose A is coordinator and B and C are participants
A asks B and C to prepare
B votes prepared
C calls B to do some work. (B is reinfected.)
B does the work and tells C it has prepared, but now it expects C to be its coordinator.
When A asks C to prepare, C propagates the request to B and votes prepared only if both B and C are prepared. (See Tree of Processes discussion later.)
Can be used to implement integrity constraint checking, triggers, and other commit-time processing, without requiring an extra phase (between phases 1 and 2 of 2PC).
Phase Zero
Suppose a participant P is caching transaction T’s updates that P needs to send to an RM (another participant) before T commits.
P must send the updates after T invokes Commit, to ensure P has all of T’s updates
P must send the updates before the RM prepares, to ensure the updates are made stable during phase one.
Thus, we need an extra phase, before phase 1.
A participant explicitly enlists for phase zero.
It doesn’t ack phase zero until it finishes flushing its cached updates to other participants.
Supported in Microsoft DTC.
7.5 Process Structuring
To support multiple RMs on multiple nodes, and minimize communication, use one transaction manager (TM) per node
TM may be in the OS (VAX/VMS, Win), the app server (IBM CICS), DBMS, or a separate product (early Tandem).
TM performs coordinator and participant roles for all transactions at its node.
TM communicates with local RMs and remote TMs.
Enlisting in a Transaction
When an Application in a transaction T first calls an RM, the RM must tell the TM it is part of T.
Called enlisting or joining the transaction
Enlisting in a Transaction (cont’d)
When an application A in a transaction T first calls an application B at another node, B must tell its local TM that the transaction has arrived.
Tree of Processes
Application calls to RMs and other applications induces a tree of processes
Each internal node is the coordinator for its descendants, and a participant to its parents.
This adds delay to two-phase commit
Optimization: flatten the tree, e.g. during phase 1
Handling Multiple Protocols
Communication managers solve the problem of handling multiple 2PC protocols by providing
a model for communication between address spaces
a wire protocol for two-phase commit
But, expect restrictions on multi-protocol interoperation.
The RM only talks to the TM-RM interface. The multi-protocol problem is solved by the TM vendor.
Complete Walkthrough
Customer Checklist
Does your DBMS support 2PC?
Does your execution environment support it? If so,
with what DBMSs?
Using what protocol(s)?
Do these protocols meet your interoperation needs?
Is the TM-DBMS interface open (for home-grown DBMSs)?
Can an operator commit/abort a blocked txn?
If so, is there automated support for reconciling mistakes?
Is there automated heuristic commit?
7.6 Three Phase Commit- The Idea
3PC prevents blocking in the absence of communications failures (unrealistic, but …). It can be made resilient to communications failures, but then it may block
3PC is much more complex than 2PC, but only marginally improves reliability — prevents some blocking situations.
3PC therefore is not used much in practice
Main idea: becoming certain and deciding to commit are separate steps.
3PC ensures that if any operational process is uncertain, then no (failed or operational) process has committed.
So, in the termination protocol, if the operational processes are all uncertain, they can decide to abort (avoids blocking).
Three Phase Commit- The Protocol
1. (Begin phase 1) Coordinator C sends Request-to-prepare to all participants
2. Participants vote Prepared or No, just like 2PC.
3. If C receives Prepared from all participants, then (begin phase 2) it sends Pre-Commit to all participants.
4. Participants wait for Abort or Pre-Commit. Participant acknowledges Pre-commit.
5. After C receives acks from all participants, or times out on some of them, it (begin third phase) sends Commit to all participants (that are up)
3PC Failure Handling
If coordinator times out before receiving Prepared from all participants, it decides to abort.
Coordinator ignores participants that don’t ack its Pre-Commit.
Participants that voted Prepared and timed out waiting for Pre-Commit or Commit use the termination protocol.
The termination protocol is where the complexity lies. (E.g. see [Bernstein, Hadzilacos, Goodman 87], Section 7.4)