-
Type: Task
-
Status: Done (View Workflow)
-
Priority: Major
-
Resolution: Done
-
Component/s: None
-
Labels:
The postgres instance on db-ics is configured with database storage in the default location, which is on the root directory. On the VMs, that is ~8G and includes /var, /tmp, and /home/pfs. As it stands we have ~5G left, ~4G useable by the db.
Once we get past the initial hardware configuration, we were planning to run centroiding on all mcs exposures. There are roughly 3600 holes; I think the rows have 48 bytes of data, plus 24 bytes of overhead plus one index. Call it 80 bytes per hole, ~300kB per exposure. We could conceivably (but will not) take ~600 frames/hour, 10 hours/night: ~1.7GB/night
There is a db LUN on the storage server for just this purpose, but I do not know how it has been provisioned, and we have not used it. Given that we are demonstrably running successfully on the current VM storage I think it is safest to simply add more. kyono has created a new 48G partition now mounted on db-ics:/data-tmp/, so we can use that.
I am not a postgres admin, but it looks trivial to move a tablespace or to create a new one on a new location. The fancier stuff quickly gets intimidating.
I propose taking a backup after tonight (saved to some other partition), and seeing how tight things really are. In any case I'll create a new tablespace on /data-tmp and be ready to shift things.