I had an issue recently on memory allocation for databases created on memsql which I raised here and found a solution. i.e: Either to delete unnecessary databases or reduce transaction-buffer system variable. So I selected 1st option to delete the unnecessary databases and to keep only 20 databases. But still they say only 64MB is allocated per a DB. So I ran the query
SHOW DATABASES EXTENDED;
on my leaf node and it surprisingly returned 130 DBs including all past dropped dbs. which causes my Alloc_durability_large to 7GB. But it should be 64MB * 20 = 1280MB.
How can I get rid of these databases and why they are not removed although I have dropped the databases in master. Do I need to drop databases in both leaf and master for it to have an effect?
And databases are replicated as _0 _1 _2 _4 eg: mark db is replicated as mark_0, mark_1 etc? should I manually delete all?
Note: I have restarted memsql still no effect.
When you run DROP DATABASE on the master aggregator, it drops the database cluster-wide, you do not need to do any operations on the leaf normally. The mark_0, mark_1, etc are partitions of the mark database on the leaves, and are normally removed when you drop database mark.
Are you running HA? It's possible you are seeing orphan databases because of operations performed while a node was offline. You can use the command CLEAR ORPHAN DATABASES to remove them - use EXPLAIN CLEAR ORPHAN DATABASES first to see what databases it thinks are orphans.