Typo3 Scheduler Jobs

TYPO3 Scheduler jobs

Source: https://github.com/dirnbauer/typo3-skills

Configure the smallest complete task set justified by the selected installation. Derive tasks, parameters, groups, and cadence from its TYPO3 14 runtime; never copy another site's rows.

Version gate

Require TYPO3 14.3.x. Stop on v13, pre-14.3 development releases, or v15 and re-evaluate the APIs. TYPO3 v14 stores task type, parameters, and execution details as structured fields; an empty tx_scheduler_task.tasktype means the v14 migration failed or remains incomplete.

Contract

1. Identify one installation and report its root, TYPO3 version, application context, and database platform before changing Scheduler data.
2. Inventory any system cron, hosting job, systemd timer, or container runner that invokes vendor/bin/typo3 scheduler:run. Configuring Scheduler records does not authorize creating, enabling, changing, or removing that external runner; automatic execution is opt-in.
3. Inventory runtime-registered task types, active Sites, storages, cache backends, installed extensions, existing groups/tasks, failures, late tasks, and v14 storage migration defects.
4. Read task-catalog.md before selecting tasks. Include every fixed Core task in the decision table, but create conditional tasks only when their applicability is proven.
5. Build the complete task list first. Then create or reuse the five functional groups below and assign every enabled task; leave no unexplained task in the ungrouped section.
6. Use the conservative baseline cadence. Change it only from measured runtime, queue depth, publishing latency, retention policy, storage change source, or an explicit operational need.
7. Set parallel execution off. Stagger expensive work, preserve a maintenance window, and run destructive jobs against verified backups and an approved retention policy.
8. Create or edit tasks through the TYPO3 v14 Scheduler form/runtime model. Do not write v12 PHP serialized task objects, copy database rows, or install a generic command/SQL scheduler helper.
9. Run each new task manually once when its effect is safe and approved, then inspect its failure state and effect. Do not invoke the global due-task runner or enable external cron by inference.
10. Report configured, conditional, rejected, and project-specific tasks with the evidence, effective interval, and execution mode for each.

Choose the execution mode explicitly

Default to manual-only unless the user explicitly requests automatic Scheduler execution. Creating recurring task records is separate from installing the infrastructure runner.

• Never install a DDEV cron add-on, write a host crontab, create a systemd/Kubernetes/hosting job, or add a deployment schedule merely because the user asked to configure Scheduler jobs.
• Audit an existing runner read-only. Change or remove it only when the user explicitly requests that operational change and the exact owner/target has been verified.
• In manual-only mode, verify that no external runner calls this installation. Leave the reviewed Scheduler records available for explicit scheduler:execute --task=<uid> use.
• In automatic mode, agree the runtime owner, working directory, logging, alerting, overlap policy, and cadence before enabling the runner. The runner cadence must cover the shortest task cadence.
• A DDEV project is not an exception. Development convenience does not imply permission for unattended execution.

Workflow: audit the selected installation

Run the bundled read-only audit from the Composer project root:

php /absolute/path/to/typo3-scheduler-jobs/scripts/audit-scheduler.php --pretty

For DDEV, stream it into the project container:

ddev exec php /dev/stdin --pretty \
  < /absolute/path/to/typo3-scheduler-jobs/scripts/audit-scheduler.php

Use --strict to return non-zero for broken v14 task storage, unregistered existing task types, or recorded execution failures. The audit redacts arbitrary task parameters and exposes only known-safe maintenance/Solr keys.

Also capture the Core view:

vendor/bin/typo3 scheduler:list
vendor/bin/typo3 scheduler:run --help
vendor/bin/typo3 scheduler:execute --help

Read discovery-and-operations.md before creating, changing, disabling, executing, or deleting a task.

Group the final task set

Create groups only after classifying all tasks. Reuse an existing group when its meaning matches.

Use task descriptions to record the applicability evidence, retention owner, selected Site/storage, and why the cadence differs from the baseline. Group disabling is an operational kill switch, not a substitute for disabling or removing an invalid task.

Apply the cadence policy

Classify frequency before choosing a cron expression:

• Very frequent: every minute or every 2 minutes. Reserve this for a measured, latency-sensitive queue. It magnifies failures and load.
• Interactive-safe middle: every 5 minutes for search/event/publishing queues; every 15–30 minutes for file discovery or extraction.
• Maintenance-safe middle: daily for low-cost privacy/status work; weekly for broad deletion, crawling, and recycler work; monthly and off-peak for measured database reorganization.
• Very infrequent: monthly, quarterly, or manual. Use this for expensive recovery and rebuild operations, not for a queue whose users expect fresh results.

Never make a task more frequent than its worst observed duration plus headroom. Keep parallel execution disabled and fix backlog/capacity before shortening the interval.

Configure the baseline

Use the detailed parameters and exceptions in task-catalog.md.

• Create daily cache GC only for selected backends that require explicit collection.
• Create daily IP anonymization for every runtime-registered IP table, with a documented retention value. Propose 30 days when no policy exists; require approval before applying it.
• Create weekly "all configured tables" GC so Core and installed extensions retain their own registered expiry periods.
• Create FAL extraction per applicable storage. Create FAL storage indexing only where files can change outside TYPO3.
• Keep database optimization conditional on MySQL/MariaDB, measured reclaimable space, and a maintenance window.
• Create conditional LinkValidator, Reports, Form, Redirects, Workspace, and extension jobs only when their installed feature and data path are in use.

Configure EXT:solr 14

If apache-solr-for-typo3/solr is installed, read solr-v14.md before writing tasks.

Create exactly one IndexQueueWorkerTask per configured TYPO3 Site that is indexed. Select the Site whose root page is the actual start of that website's records; EXT:solr stores this root page UID. Default to every 5 minutes and 50 documents per run, then tune from queue age and measured duration. Add the Event Queue Worker only for delayed monitoring (monitoringType = 1).

Do not schedule recurring forced reindexing or Solr optimize as routine maintenance. Use them as reviewed recovery/maintenance operations. EXT:solr 14 removed the forced-webroot task field; never recreate it from older installations.

Verify and hand off

1. Run scheduler:list; verify every enabled task has a group, expected frequency, and no failure. A future next execution describes the task schedule; it does not prove automatic execution.
2. Execute each new task by UID with scheduler:execute --task=<uid> in a safe environment.
3. Verify lastexecution_time, next execution, logs, affected rows/files, Solr queue age, and notification delivery as applicable. Run global scheduler:run only when the user explicitly asks to simulate or operate the due-task runner.
4. In automatic mode, confirm the external runner independently. In manual-only mode, confirm its absence and provide the exact explicit command for each task.
5. Re-audit after one full weekly cycle only in automatic mode, and after every TYPO3/extension major update in either mode.

Deliver a table with task/group, task type, parameters, cadence, applicability evidence, manual-run result, execution mode, and keep/change/remove verdict. Never claim automatic operation while a required task is late, failed, ungrouped, or dependent on an unverified runner. A manual-only configuration may be complete when the runner is intentionally absent and that limitation is reported prominently.