Description of problem: Running heat-manage purge_deleted deletes only part of the data, keeping a lot of useless data (heat.raw_template). MariaDB [heat]> SELECT table_name AS `Table`, round(((data_length + index_length) / 1024 / 1024), 2) `Size in MB` FROM information_schema.TABLES WHERE table_schema = "heat" AND table_name = "raw_template"; +--------------+------------+ | Table | Size in MB | +--------------+------------+ | raw_template | 3240.69 | +--------------+------------+ #heat-manage purge_deleted MariaDB [heat]> SELECT table_name AS `Table`, round(((data_length + index_length) / 1024 / 1024), 2) `Size in MB` FROM information_schema.TABLES WHERE table_schema = "heat" AND table_name = "raw_template"; +--------------+------------+ | Table | Size in MB | +--------------+------------+ | raw_template | 1816.08 | +--------------+------------+ Version-Release number of selected component (if applicable): openstack-heat-api.noarch 2015.1.2-9.el7ost openstack-heat-api-cfn.noarch 2015.1.2-9.el7ost openstack-heat-api-cloudwatch.noarch 2015.1.2-9.el7ost openstack-heat-common.noarch 2015.1.2-9.el7ost openstack-heat-engine.noarch 2015.1.2-9.el7ost openstack-heat-templates.noarch 0-0.8.20150605git.el7ost How reproducible: Steps to Reproduce: 1. heat-manage purge_deleted 2. 3. Actual results: keeps some data Expected results: purge all, raw_template keeps thousands of entries Additional info:
this issue would be within the heat application itself, which is responsible for emitting the appropriate SQL to the database in order to delete the items. Only if it could be shown that this issue reproduces *only* when galera is in play, and not when plain mariadb is in use, would that even suggest the possibility of the issue being part of mariadb-galera and even then it still would likely be an issue in heat. If the purge_deleted command has issues there's likely an upstream issue for heat in launchpad already.
better component
By default (i.e. with no arguments passed to heat-manage, as indicated in the description) only stacks that were deleted more than 90 days ago are purged. Are you certain that every stack was deleted more than 90 days ago? If not then completely clearing the table is definitely not the expected result. Have you tried with a shorter horizon to see if that removes more data?
Hi, thx Zane, that did the trick, my bad i must have overlooked the flag, sorry.