Repository navigation
Conversation
hajiler
force-pushed
the
shared-volumes/pr3-provisioned-external-volumes
branch
2 times, most recently
from
November 9, 2026 22:17
c49c5b2 to
177fdb4
Compare
Collaborator
Author
|
Rebased onto the current #3744 head (adds its apitool example/annotation fix); the last commit is still the one to review. |
41 tasks
1 of 2 tasks
An ExternalVolume is an atespace-scoped record of a CSI volume that already exists in the storage system: the driver, the volume ID and the volume context the node plugin needs to mount it. Its lifetime is independent of any actor, and an atespace cannot be deleted while it holds one. This adds the resource, its Postgres table and store methods, the OpenFGA type and RPC permissions, the Create, Get, List and Delete RPCs, and the kubectl-ate commands. Create requires the ExternalVolumes preview gate and a global owner. Get, List and Delete work with the gate off, so turning it off cannot strand a record. Delete removes the record and leaves the storage in place. Nothing references an ExternalVolume yet. Actor templates still provision a volume per actor from external.template. Signed-off-by: Elijah Rodriguez-Beltran <elijahrb@google.com>
Every volume provisioned for an actor from external.template now has an ExternalVolume in the actor's atespace, named in the actor's ExternalVolumeStatus.external_volume_name. The record is written right after CreateVolume returns and deleted by the actor's delete workflow after the storage, so a retry after a crash in between converges: the retried CreateVolume returns the same volume and the record it left is reused when it names that volume. The actor status keeps its copy of the volume handle, so resume, suspend and delete read nothing new. The record name is the volume name with a hash of the actor UID and volume name, which keeps it a DNS label and unique across actors and across a namesake created after the actor is deleted. Actors provisioned before this change have no record, and deleting them tolerates that. Signed-off-by: Elijah Rodriguez-Beltran <elijahrb@google.com>
hajiler
force-pushed
the
shared-volumes/pr3-provisioned-external-volumes
branch
from
November 11, 2026 22:54
177fdb4 to
28ce447
Compare
Cover a failed storage delete keeping the record, a volume already gone from storage still losing its record, and store errors while recording or deleting one. Validate external_volume_name's format and update rules. The partial-failure functional test checks that only the provisioned volume gets a record and that none is left after delete. Rename the maxPrefix variable in provisionedExternalVolumeName, which shadowed the builtin max. Note in the resource model that Substrate writes ExternalVolumes too.
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Part of #1637. Stacked on #3744: the first commit is that PR, review the second (
ateapi: record each actor-provisioned volume as an ExternalVolume). I will rebase ontomainonce #3744 merges.Moves the per-actor volume bookkeeping onto the
ExternalVolumeresource without changing what an actor mounts: still one volume per actor, provisioned fromexternal.template, deleted with the actor.ExternalVolumein the actor's atespace right afterCreateVolumereturns, and names it in the newstatus.externalVolumes.volumes[].externalVolumeName. The status keeps its copy of the handle, so the hot paths read nothing new.FAILED_PRECONDITION.<volume>-<hash of actor UID and volume>: deterministic, unique across actors and recreations, and within a DNS label whatever the volume name's length.Known gap until actor references land:
DeleteExternalVolumedoes not yet refuse a record a live actor provisioned. The actor keeps working from its status copy and its delete still removes the storage.AI use: drafted with an AI agent under my direction.
hack/verify-all.sh,apitool validate,hack/verify/proto-breaking.sh. The e2e sub-testcombinedvolumes/ExternalVolumeRecordedis new and has not been run on a cluster yet.