daqd raw minute trends should require less manual maintenance for cymacs
The raw minute trends take manual management to deal with disk usage. This is due to the raw trends having no end period for the data. It just writes to the file.
In production when we run low on disk we:
- stop the daqd
- rename the raw frame directory to a temp name
- reconfigure daqd to read from the new temp name
- restart daqd
- move (and verify) the frames from the temp area to a large archival area
- reconfigure daqd to read from the newly backed up trends from the archival area
- restart daqd again
- delete the newly backed up trends
This is a bit much to ask for a simple test setup.
We should:
- Allow the daqd to have a period/timespan for the raw trends (1 month, 1 week, 1 day).
- Configure a directory that it will put these items under
- raw trends would be put under the directory ( <path> / start time / hash directory / raw trend files )
- As the period is exceeded a new 'start time' directory is used.
- The old directories (and all other directories under the path) get automatically added to the list of archive paths for the raw trends.
- now a frame wiper can be written which removes old trends, but does not touch current trends.