Recovery

As explained in Think Distributed Systems

Recovery is a technique for restoring a system or transaction to an acceptable state after failure or interruption.

A recovery strategy restores a system or transaction from an illegal state to a legal, acceptable state after failure. It applies only where contingencies exist; intolerable states cannot be recovered from. section 3.1

A transaction recovery strategy typically:

• Examines persistent records after restart to find transactions that were not committed. section 5.3.5
• Applies Undo operations, which must be restartable and idempotent because crashes can interrupt or repeat them. section 5.3.5
• Uses the recorded transaction state to choose abort, commit inquiry, REDO, or UNDO after a participant failure. section 6.3.2

A participant recovers

After a distributed transaction participant restarts, a pre-vote failure leads to persistent abort; a recorded vote-to-commit requires asking the coordinator; a recorded commit requires REDO; a recorded abort requires UNDO. section 6.3.2

Figure 3.1 System states and state transitions

Use the state-transition diagram to trace how failures move a system from an illegal state toward a legal state. section 3.1

How Think Distributed Systems covers it

How deep it goes

Explains restart logs, idempotent Undo, and state-dependent REDO or UNDO, while showing that coordinator failure can block progress.

Examples

Uses crash restart, Transaction Undo Logs, participant failures, coordinator failure, and state-specific recovery actions.

What it assumes

Assumes transaction logs and distributed transaction roles, while defining failure recovery and its limits from first principles.

What you can do after

Diagnose/debug with it by tracing persistent transaction state and selecting restartable Undo, REDO, abort, or coordinator inquiry.

How it connects

Connects failure recovery with failure tolerance, transactions, transaction logs, REDO, UNDO, and two-phase commit.

liveBookmanning.com

Read Think Distributed Systems on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Akka in Action, Second EditionA recovery strategy is a planned response that restores operation or limits the effects of a fault or failure.
• 

  liveBookmanning.com
  Elixir in Action, Third EditionError detection and recovery is a set of mechanisms for identifying failures and restoring affected computation or service.

Compare how 3 books explain recovery →

Related concepts

• incident responsepart of · Incident response is the organized process of detecting, managing, investigating, containing, and recovering from operational or security incidents.
• journaluses · A journal is an append-oriented, sequential record of events or changes retained for later retrieval, review, recovery, or replication.

This explainer was generated by Manning's knowledge graph from the text of Think Distributed Systems. The section links point to the parts of the book it draws on.