Description of problem: There is an information gap when running the DB's cleanup for Nova services before running a Major Upgrade on production environments. Currently the command we run is: nova-manage db archive_deleted_rows --max_rows 100 --until-complete Still, the archived rows are stored in shadow tables so we still have a big DB, also there is some specs to remove them but nothing landed/usable. This spec shows basically the problem: https://blueprints.launchpad.net/nova/+spec/purge-deleted-instances-cmd - Can you clarify the process of cleaning and maintain Nova databases?
(In reply to Carlos Camacho from comment #0) > - Can you clarify the process of cleaning and maintain Nova databases? I'm going to interpret this as a request to implement purge-deleted-instances, because if you just want information you're better asking for it on a mailing list.
Hi Matthew, Yeah, well we will like first to know what to check before asking for the purge-delete-instances implementation. Let's keep this BZ to track all the effort to document/implement this and I'll send an email to the mailing list. Thanks, Carlos.
Ok, This is the mail thread in the mailing list[1]. Even though would be nice if you can give us information about it or reply to the thread. [1]: http://lists.openstack.org/pipermail/openstack-dev/2018-February/126914.html Thanks, Carlos.
This has nothing to do with fast-forward upgrade that I know of, other than potentially pre-purging the database to make schema changes faster. The general process here is to archive the db per the command you mentioned, followed by a purge of the shadow tables, which currently is only doable in SQL: DELETE FROM shadow_instances; for example. I do think we should make this an RFE and implement it though, for what its worth.
I'm closing this as a dupe of bz1154875, which is new DB purging functionality merged in master and to be released in OSP14. *** This bug has been marked as a duplicate of bug 1154875 ***