I have MemSQL 6.7.12, with a Columnar database with 10 tables.
In MemSQL Studio it shows this database disk usage is of only 279MB, but in my cluster actually, the disk usage is of over 12GB!. Am i missing something? It is possible to free up this space?
MemSQL pre-allocates log files for each partition. That can take up a fixed amount of space on an otherwise small database. Other space is taken for data files, plan cache, snapshots, backups and so on. To see the file sizes, you can find out where the data directory is, like so:
memsql> show variables like '%datadir%';
+---------------+-----------------------------------------------------------+
| Variable_name | Value |
+---------------+-----------------------------------------------------------+
| datadir | /var/lib/memsql/593f03e0-e5b2-4ef5-bf58-ddccf8af69c2/data |
+---------------+-----------------------------------------------------------+
Then cd to that directory on a leaf node and do, for example,
Thanks for your reply, indeed I can see my leaf node directories; and actually logs, plan cache and snapshots directories doesn’t take much space. (even set plan expiration limits to low values)
The data files that are under each leaf data directory (…/columns/…) are the ones that takes much more than what MemSQL Studio says as disk usage.
Any other place I should look into?