You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
So that's a duplication. For example PUT /workflow/{identifier} takes a JSON workflow and if identifier != workflows["graph"]["id"] that we get a 400 error.
The id in the graph attribute is used for logging and ewoks events.
The file names are used to read, delete and list workflows.
These are no equal and the scope is the ewoksid29 package. Neither is guaranteed to be unique in the scope of ewoksserver.
So the question is, what would be the id in the scope of ewoksserver?
Should we change the way we handle workflow id's in ewoksserver overall and decouple them from graph id's? That could lead to confusion when debugging. Also the id's in the ewoks events need to be the same as the workflow id's in the server I think.
Ewoksserver could load the workflows instead of letting the worker load them. In that case it could change workflows["graph"]["id"] when executing to the ewoksserver id. To be discussed.
Activity
poautran commentedon Jan 10, 2026
In GitLab by @woutdenolf on Jan 10, 2026, 17:23 GMT+1:
About current workflow identifiers in ewoksserver:
id.So that's a duplication. For example
PUT /workflow/{identifier}takes a JSON workflow and ifidentifier != workflows["graph"]["id"]that we get a 400 error.The
idin the graph attribute is used for logging and ewoks events.The file names are used to read, delete and list workflows.
poautran commentedon Jan 10, 2026
In GitLab by @woutdenolf on Jan 10, 2026, 17:35 GMT+1:
The workflow
ewoksid29.workflows.ambigator_runhasambigator_run.jsonworkflows["graph"]["id"] = workflow_ID29ESRF_ambigator_runThese are no equal and the scope is the
ewoksid29package. Neither is guaranteed to be unique in the scope ofewoksserver.So the question is, what would be the id in the scope of ewoksserver?
Should we change the way we handle workflow id's in ewoksserver overall and decouple them from graph id's? That could lead to confusion when debugging. Also the id's in the ewoks events need to be the same as the workflow id's in the server I think.
Ewoksserver could load the workflows instead of letting the worker load them. In that case it could change
workflows["graph"]["id"]when executing to the ewoksserver id. To be discussed.woutdenolf commentedon Jul 6, 2026
Discussed with @loichuder : in the first approach everything will be handled in the backend
woutdenolf commentedon Jul 6, 2026
Needs workflow discovery to be exposed by the workers: ewoks-kit/ewoksjob#341.