Skip to content

restart.md

Latest commit

 

History

History
104 lines (78 loc) · 2.86 KB

restart.md

File metadata and controls

104 lines (78 loc) · 2.86 KB

Restart Semantics

Lemric BatchProcessing supports full restart semantics — a failed or stopped job can be re-launched from its last checkpoint.

How Restart Works

  1. A JobInstance is identified by (jobName, hash(identifyingParameters)).
  2. Re-running with the same identifying parameters finds the existing JobInstance:
Existing State Behaviour
COMPLETED JobInstanceAlreadyCompleteException is thrown
STARTED JobExecutionAlreadyRunningException is thrown
FAILED A new JobExecution is created — job resumes
STOPPED A new JobExecution is created — job resumes
ABANDONED JobRestartException is thrown
  1. On restart each step checks for a previous StepExecution:
    • If the step was COMPLETED, it is skipped (unless allowStartIfComplete is set).
    • If the step was FAILED or STOPPED, it re-executes with the persisted ExecutionContext.

ExecutionContext & Checkpoints

The ExecutionContext is persisted after every committed chunk. Readers and writers implementing ItemStreamInterface store their cursor position:

public function open(ExecutionContext $ctx): void
{
    // Restore from last checkpoint
    $this->currentLine = $ctx->getInt('reader.current.line', 0);
}

public function update(ExecutionContext $ctx): void
{
    // Save checkpoint
    $ctx->put('reader.current.line', $this->currentLine);
}

Preventing Restart

$job = $jobBuilderFactory->get('oneTimeJob')
    ->start($step)
    ->preventRestart(true)
    ->build();

Re-Running Completed Steps

By default, completed steps are skipped on restart. To force re-execution:

$step = $stepBuilderFactory->get('alwaysRunStep')
    ->chunk(100, $reader, $processor, $writer)
    ->allowStartIfComplete(true)
    ->build();

Stopping a Running Job

Request a graceful stop via the operator:

$operator->stop($executionId); // returns bool

The framework checks the interruption policy before each chunk. The job will complete the current chunk, persist the context, and exit with STOPPED status.

Abandoning a Job

Mark a stopped/failed job as abandoned to permanently prevent restart:

$operator->abandon($executionId); // returns the updated JobExecution

Signal Handling

When ext-pcntl is available, SignalHandler / SignalJobInterruptionPolicy intercept SIGTERM and SIGINT signals and set the job to STOPPING for a graceful shutdown.

Next Steps