mudhorn/n8n
1
0
Fork 0
mirror of https://github.com/n8n-io/n8n.git synced 2024-12-25 20:54:07 -08:00
Code Issues Actions Packages Projects Releases Wiki Activity

refactor(core): Add an option to use simple recovery process by default (#7097)

Browse source
This commit is contained in:
कारतोफ्फेलस्क्रिप्ट™ 2023-09-04 19:58:36 +02:00 committed by GitHub
parent ee36f2d20b
commit cffda65b33
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 22 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

6
packages/cli/src/config/schema.ts
View file

@ -1124,6 +1124,12 @@ export const schema = {
env: 'N8N_EVENTBUS_LOGWRITER_LOGBASENAME',
},
},
crashRecoveryMode: {
doc: 'Should n8n try to recover execution details after a crash, or just mark pending executions as crashed',
format: ['simple', 'extensive'] as const,
default: 'extensive',
env: 'N8N_EVENTBUS_RECOVERY_MODE',
},
},
redis: {

10
packages/cli/src/databases/repositories/execution.repository.ts
View file

@ -216,6 +216,16 @@ export class ExecutionRepository extends Repository<ExecutionEntity> {
return newExecution;
}
async markAsCrashed(executionIds: string[]) {
await this.update(
{ id: In(executionIds) },
{
status: 'crashed',
stoppedAt: new Date(),
},
);
}
async updateExistingExecution(executionId: string, execution: Partial<IExecutionResponse>) {
// Se isolate startedAt because it must be set when the execution starts and should never change.
// So we prevent updating it, if it's sent (it usually is and causes problems to executions that

15
packages/cli/src/eventbus/MessageEventBus/MessageEventBus.ts
View file

@ -126,18 +126,15 @@ export class MessageEventBus extends EventEmitter {
LoggerProxy.info(` - ${workflowData.name} (ID: ${workflowData.id})`);
}
}
if (this.logWriter?.isRecoveryProcessRunning()) {
const recoveryAlreadyAttempted = this.logWriter?.isRecoveryProcessRunning();
if (recoveryAlreadyAttempted || config.getEnv('eventBus.crashRecoveryMode') === 'simple') {
await Container.get(ExecutionRepository).markAsCrashed(unfinishedExecutionIds);
// if we end up here, it means that the previous recovery process did not finish
// a possible reason would be that recreating the workflow data itself caused e.g an OOM error
// in that case, we do not want to retry the recovery process, but rather mark the executions as crashed
LoggerProxy.warn('Skipping recover process since it previously failed.');
for (const executionId of unfinishedExecutionIds) {
LoggerProxy.info(`Setting status of execution ${executionId} to crashed`);
await Container.get(ExecutionRepository).updateExistingExecution(executionId, {
status: 'crashed',
stoppedAt: new Date(),
});
}
if (recoveryAlreadyAttempted)
LoggerProxy.warn('Skipped recovery process since it previously failed.');
} else {
// start actual recovery process and write recovery process flag file
this.logWriter?.startRecoveryProcess();