The stored procedure
COMPACT_METADATA_TABLES reclaims unused space
from the database where the Virtual DataPort metadata is stored.
Virtual DataPort stores its metadata in a local database. For each virtual database you create, it creates in this database a set of tables that store the metadata: one table for the data sources, another one for the views, another for the web services, etc. Over time, as the users create and delete elements, the space used by the metadata database grows because the database does not return unused space to the operating system.
Run this procedure when the size of the directory
<DENODO_HOME>/metadata/db/dbmetadata grows over a few gigabytes.
This procedure compacts the tables of the metadata database one after the other; not all of them at once.
Take into account the following:
When compacting a table, it acquires an exclusive lock on that table. This lock prevents Virtual DataPort from reading or writing data from the table. This lock is relased once the table is compacted.
It is a memory-intensive process.
Depending on the “mode” selected (see the parameter
modebelow), it creates temporary files. This is a problem if the disk is already full.
Therefore, try to run this procedure at a time the load of the server is very low or there is no load. If the server belongs to a cluster of Denodo servers, remove it from the cluster to avoid affecting external applications that connect to Denodo.
Before using this procedure for the first time, execute this VQL statement using an administrator account:
CREATE PROCEDURE compact_metadata_tables CLASSNAME = 'com.denodo.vdb.contrib.storedprocedure.CompactMetatadataTablesProcedure';
COMPACT_METADATA_TABLES( compaction_type : int )
compaction_type: mode used to compact the tables. This procedure compacts each table of the metadata database one after the other, not all of them at once.
These are the possible values:
0(no sequential): with this mode, for each table of the metadata database, the procedure compacts the table and at the same time all its indexes.
1(sequential): with this mode, for each table of the metadata database, the procedure compacts the table and once it finishes compacting the table, it compacts its indexes one by one.
2(in place): this mode does not use temporary tables and indexes. It just moves rows around the same space.
1(sequential) are guaranteed to recover the maximum amount of free space that the mode
2(in place), at the cost of temporarily creating new tables and indexes. This is because with the modes
1, the database copies active rows to newly allocated space and once it finishes copying these rows, it releases the space used by the old table. With mode
2, all the work is done in the same space.
0uses more memory and disk space than mode
1. To compact an index, a temporary file is created. With the mode
0all the indexes of a table are compacted simultaneously so the temporary files for those indexes are created at the same time. Whereas with mode
1, the indexes of a table are compacted one by one and therefore, at all times there is only one temporary file for indexes.
On the other hand, the mode
0is much faster than mode
For the same reason, mode
0is more CPU intensive than mode
2cannot guarantee it will recover all available space, but if the hard drive is full, you make sure that the space temporary used by the database does not increase.
The output schema of this procedure has the following fields:
COMPACTION_TYPE_NAME: name of the mode used to compact. That is, the name of the mode entered as a parameter:
METADATA_TABLES_COMPACTED: number of tables of the metadata database that have been compacted.
ERRORS: this database invokes the same procedure over each table of the metadata database. This field will hold the number of times the execution of this procedure failed.