📋 Documentation Version Notice

You're reading the documentation for a development version (master). For the latest released version, please have a look at v1.7.

Detached BIOMERO Workflows

New in version 1.8.0: Detached workflow execution is controlled by an opt-in feature flag. The fresh NL-BIOMERO demo configuration enables it, but existing and custom deployments continue to use inline execution until BIOMERO_DETACHED_WORKFLOWS is explicitly enabled.

Note

Summary for system administrators:

Enable detached analysis with BIOMERO_DETACHED_WORKFLOWS=TRUE.
Detached workflows continue when the submitting OMERO.web session ends.
When detached mode is enabled, the user’s OMERO session does not need to cover the complete Slurm runtime. The OMERO script timeout must still accommodate long server-side transfer and import operations.
The setting applies to BIOMERO analysis workflows, not browser uploads.

Overview

BIOMERO analysis may include data transfer, format conversion, Slurm queueing and execution, and result import. These operations can take considerably longer than a normal OMERO.web session.

Detached mode hands an accepted analysis request to biomeroworker for background execution. Progress and results remain available through BIOMERO after the submitting browser or OMERO session has ended.

When detached mode is disabled, the workflow runs inline. The OMERO script and requesting session must then remain available until processing and result import have completed. Inline deployments may require longer OMERO timeouts for long-running workflows.

Detached mode covers workflows launched through SLURM_Run_Workflow.py and SLURM_Run_Workflow_Batched.py. It does not detach an upload that is still being transferred from a browser.

Configuration

Set the following value in the NL-BIOMERO .env file:

BIOMERO_DETACHED_WORKFLOWS=TRUE

The supplied Compose files pass this feature flag to the required services. If the flag is absent or FALSE, workflows use inline execution.

Important

Update biomeroworker to the corresponding NL-BIOMERO release before enabling detached workflows. The worker must include the same compatible biomero-scripts release as OMERO.server. The supplied NL-BIOMERO images keep these versions aligned through BIOMERO_SCRIPTS_VERSION.

Detached mode currently supports one active biomeroworker per BIOMERO tracking database. Multiple worker replicas are not supported.

The background worker can be tuned with these optional settings:

Variable	Default	Description
`BIOMERO_MAX_ACTIVE_WORKFLOWS`	`4`	Maximum number of workflows managed concurrently by the worker.
`BIOMERO_SUPERVISOR_POLL_SECONDS`	`10`	Interval between checks for newly submitted workflows.
`BIOMERO_SUPERVISOR_STARTUP_GRACE_SECONDS`	`60`	Delay before a restarted worker resumes unfinished workflows.

Increasing BIOMERO_MAX_ACTIVE_WORKFLOWS does not increase Slurm capacity. For example, if two workflows request a GPU and only one GPU is available, Slurm runs one job while the other remains pending.

Apply a change to the detached-mode setting by recreating the affected services:

docker compose up -d --force-recreate biomeroworker omeroweb

Session and timeout settings

Detached workflows are designed to run with ordinary user-session settings. The supplied environment files keep user sessions short while allowing server-side scripts to complete long transfer and import operations:

OMERO_SCRIPTS_TIMEOUT=604800000
OMERO_SESSIONS_TIMEOUT=600000
OMERO_WEB_SESSION_COOKIE_AGE=86400

These values correspond to a seven-day OMERO script timeout, a ten-minute OMERO session idle timeout, and a one-day OMERO.web cookie age. The long script timeout accommodates operations such as importer polling, which may continue for several hours. It does not extend the submitting user’s session. These values remain configurable in .env according to local security and access policies.

Important

The short session timeout is suitable for long workflows only when detached mode is enabled. In inline mode, the requesting session and top-level script remain involved until result import completes; long workflows may therefore require a longer session timeout as well.

OMERO_WEB_SESSION_EXPIRE_AT_BROWSER_CLOSE controls browser cookie behavior and may be set independently. Detached analysis does not require this setting to be disabled.

Normal timeouts must still allow request validation and submission to finish. They must also be appropriate for individual OMERO-side operations such as data transfer and result import. They do not need to cover time spent waiting in the Slurm queue or running the analysis job.

Operational behavior

Submitted workflows appear in the BIOMERO workflow overview, where their progress and final status can be monitored independently of the original web session.

Workflow state is stored in the BIOMERO tracking database. If biomeroworker restarts, it resumes unfinished detached workflows and continues monitoring jobs that have already been submitted to Slurm.

Do not disable detached mode while workflows are queued or running. They remain recorded but cannot progress until the background worker is available again.

Confirm the active configuration with:

docker compose exec biomeroworker printenv BIOMERO_DETACHED_WORKFLOWS
docker compose exec omeroweb printenv BIOMERO_DETACHED_WORKFLOWS

Both services should report the same value.

Troubleshooting

A workflow remains queued: Confirm that biomeroworker is running and can access the BIOMERO tracking database, OMERO server, shared storage, and Slurm cluster. Review the worker logs for connection or permission errors.
A Slurm job remains pending: Inspect the Slurm pending reason. Available partitions, GPUs, memory, priorities, and cluster limits determine when a submitted job starts.
Workflow progress is not updating: Confirm that the worker is running and that both the worker and OMERO.web connect to the same BIOMERO tracking database.