)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":4523,"name":"Eric Harney","email":"eharney@redhat.com","username":"eharney"},"change_message_id":"320489c34b30885aae2b8e0a73af4b909ae9bac3","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"d2119a31_18aaab7f","updated":"2026-10-16 17:33:12.000000000","message":"I think this needs a flag at the API layer to allow this behavior -- it has implications for quotas, billing, etc. from growing the dest volume.  (The check here I think is also not 100% accurate for all drivers.)","commit_id":"af511c4a9fdd05458c688e80fc3646dc19439c92"},{"author":{"_account_id":32861,"name":"Jan Wasilewski","email":"jan.wasilewski@atman.pl","username":"janwasilewski"},"change_message_id":"f7bc61c368386be927627617df2969978d331a33","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":3,"id":"3c956d38_f9fc5b08","in_reply_to":"d2119a31_18aaab7f","updated":"2026-10-19 12:38:38.000000000","message":"Thanks for the review, Eric. Makes sense that growing the volume should not happen implicitly.\n\nProposal:\n- New microversion (3.72) adding an optional boolean to os-retype, e.g. \"allow_size_increase\" (name open for suggestions).\n- When an available, unencrypted volume is retyped to a type with front-end LUKS encryption and the flag is set, the API reserves the extra 1 GB together with the existing retype reservations and passes the new size down to the generic migration (volume RPC API bump).\n- Without the flag (or on older microversions) the behavior stays as it is today. Optionally we could fail fast in the API with a clear message instead of the late ENOSPC in dd, if you think that is acceptable for older microversions.\n- I\u0027d narrow the check to control_location front-end and LUKS providers (luks/luks2). Which drivers did you have in mind where it is not accurate - RemoteFS (qcow2/luks files), RBD native encryption, back-end encryption? I can add a driver-level way to opt out if needed.\n\nDoes this need a spec, or is a blueprint + this change enough? I\u0027ll also bring it up at Wednesday\u0027s meeting.","commit_id":"af511c4a9fdd05458c688e80fc3646dc19439c92"},{"author":{"_account_id":32861,"name":"Jan Wasilewski","email":"jan.wasilewski@atman.pl","username":"janwasilewski"},"change_message_id":"020e1e6d54791d34b3a8c186000a9bb85139aeaf","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":4,"id":"0d7fc4b2_0bc92ee3","in_reply_to":"3c956d38_f9fc5b08","updated":"2026-10-22 10:56:33.000000000","message":"As agreed at the Cinder meeting on 2026-10-21, I proposed a spec for the opt-in flag: https://review.opendev.org/c/openstack/cinder-specs/+/1011834 (blueprint retype-encrypt-allow-size-increase).\n\nIt also covers the driver accuracy point: RBD and the RemoteFS based drivers allocate the LUKS header in addition to the volume size, so the spec adds a backend capability and grows the volume only where the header lives inside it.\n\nI set this change to WIP until the spec is reviewed. Patch set 4 only updates the commit message (spec link, Assisted-By disclosure).","commit_id":"d9d960736c67fa694f0c9cc4a9d9bb93d803fef4"}]}