Details
-
Story
-
Status: Done
-
Resolution: Done
-
None
-
Architecture
-
No
Description
A butler pointing at a registry fails to see newly ingested data. This is a major nuisance when observing as you are always interested in the data that you have just taken, and thus have to restart the butler after every exposure.
Attachments
Activity
and is NFS involved? The SQLite FAQ indicates that everything should work so long as OODS is writing to a local file (not using NFS) and other butler users aren't ever writing to the sqlite file.
This is due to a misfeature in the code that explicitly avoids sharing a SQLite file (which should work over NFSv4).
With Gen3 rapidly approaching, the hope is that this is not needed for very long and so any performance impacts for large-scale ingestion (which I don't think are needed in the Gen2 world) will not be noticeable.
Looks like a nice cleanup. The tempfile each time completely explains the problem.
Is this something to do with how shared sqlite database work when two processes are attached?
(why is this a pipe_tasks issue and not a daf_persistence issue?)