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
{{ message }}
Repository navigation
Import: re-check an unknown import outcome once Studio can read the job again #1957
After #1955, an import that Studio stops following ends as Outcome unknown. That covers five failed get_job polls in a row, the 10-minute deadline, and a job the polled node doesn't have. Studio says the import may still be running or may have finished, and points to the table and the Jobs tab.
One case in there is a real failure that Studio reports as unknown. The job reaches ERROR on the server between Studio's last good poll and the five failed ones. Studio never reads the ERROR, so it can't call the import failed. That part is by design: Studio can't report an outcome it never observed. #1955 records it in src/features/instance/databases/DESIGN.md.
Possible improvement
Once Studio can reach the instance again, it could check that job once more and replace the unknown outcome with the real one. It would say that the import completed, or that it failed with the job's message. That could happen the next time the import dialog or table is opened, or on a slow background check while the page stays open.
The catch: get_job reads only the node it reaches, because hdb_job doesn't replicate. So on a cluster, a recheck can miss the job just as the original polls did. A recheck that doesn't find the job should leave the outcome unknown, never turn it into a failure.
Low priority. The current outcome is honest, and the user is already told to check the table before importing again.
Activity