Glance db purge utility

https://blueprints.launchpad.net/glance/+spec/database-purge

This spec adds the ability to sanely and safely purge deleted rows from the glance database for all relevant tables. Presently, we keep all deleted rows. I believe this is unmaintainable as we move towards more upgradable releases. Today, most operators depend on manual DB queries to delete this data, but this exposes the DB to human errors.

The goal is to have this be an extension to the glance-manage db command. Similar specs are being submitted to all the various projects that touch a database.

Problem description

Very long lived OpenStack installations will carry around database rows for years and years. This brings following problems:

  • If deleted data is kept in the DB, the number of rows can grow very large taking up the disk space of nodes. Larger disk space means more worry for disaster recovery, long running non-differential backups, etc.
  • Large number of deleted rows also means, an admin or authorized owner querying for the corresponding rows will get 5xx responses timing out on the DB, eventually slowing down other queries and API performance.
  • DB upgradeability is a big challenge if the older data style are less or inconsistent with the latest formats. An example would be the image locations string where older location string styles are different from the latest.

To date, there is no “mechanism” to programmatically purge the deleted data. The archive rows feature doesn’t solve this.

Proposed change

The proposal is to add a “purge” method to DbCommands in glance/glance/cmd/manage.py. This will take a maximum count of rows to delete and number of days argument and use that for a data_sub match. Like:

DELETE FROM images
  WHERE deleted != 0 AND deleted_at > data_sub(NOW()...)
  LIMIT count;

Alternatives

Today, this can be accomplished manually with SQL commands, or via script. There is also the archive_deleted_rows method. However, this won’t satisfy certain data destruction policies that may exist at some companies.

Data model impact

None, all tables presently include a “deleted_at” column.

REST API impact

None, this would be run from glance-manage

Security impact

None, This only touches already deleted rows.

Notifications impact

None

Other end user impact

This affects the operator’s ability to identify and delete all those rows from the DB that have location stored for backend data entities, that are not deleted (or partially deleted) due to some error at the time of the DELETE api call or during scrubber’s data purge run.

This has another operator/user impact for the calls that require changes-since filter. It is required by the Nova proxy API and exists in v1 (and potentially in v2). Purging the deleted data results into information evaporation that the changes-since filter use case is designed for. One example, if it is needed to check an old snapshot that has been deleted but has some vital info for cross checking. Some operators give/want to give that functionality to users. Purging breaks that contract and operators need to be aware about it.

Performance Impact

This has the potential to improve performance for very large databases. Very long-lived installations can suffer from inefficient operations on large tables.

Another performance impact is on the negative side unless the operators are being careful about the count. A long running purging operation will potentially cause delay in the upgrade process. It is also likely to result into a different differential backup of the DB and that may delay the process.

Other deployer impact

Some developers who want to optimize calls like re-adding a deleted image-member to image while re-sharing that image, will have to consider the rows can be deleted from now on.

Developer impact

None

Implementation

Assignee(s)

Primary assignee:
mmagr

Reviewers

Core reviewer(s):
flaper87 nikhil-komawar
Other reviewer(s):
None

Work Items

Add purge functionality to manage.py db/api.py db/sqlalchemy/api.py Add tests to confirm functionality Add documentation of feature

Dependencies

None

Testing

The test will be written as such. Three rows will be inserted into a test db. Two will be “deleted=1”, one will be “deleted=0” One of the deleted rows will have “deleted_at” be NOW(), the other will be “deleted_at” a few days ago, lets say 10. The test will call the new function with the argument of “7”, to verify that only the row that was deleted at 10 days ago will be purged. The two other rows should remain.

Documentation Impact

The documentation needs to emphasize that the image_locations table will be trimmed, which will destroy all information about where the image was stored in various backends. The operator should keep this in mind when selecting the number-of-days value for the purge function.

References

This was discussed on both the openstack-operators mailing list and the openstack-developers mailing lists with positive feedback from the group.

http://lists.openstack.org/pipermail/openstack-dev/2014-October/049616.html