brandon/nas-burnin
Watch
1
0
Fork
You've already forked nas-burnin
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

truenas: migrate client from deprecated REST to JSON-RPC 2.0 over WebSocket #1

Merged
brandon merged 1 commit from truenas-jsonrpc-migration into main 2026-07-14 00:12:28 -04:00
Conversation 0 Commits 1 Files changed 4 +544 -77
brandon commented 2026-07-14 00:12:19 -04:00
Owner
Copy link

TrueNAS removes the /api/v2.0 REST API in 26.04. REST was stateless, so the
12s poll loop re-authenticated on every request (~130k auths/day, tripping the
deprecation alarm). Switch app/truenas.py to JSON-RPC 2.0 over a single
persistent, authenticated wss://.../api/current connection: requests are
multiplexed by JSON-RPC id and demuxed by one background reader task, so auth
happens once per connection instead of per request.

Method mappings verified against a live TrueNAS SCALE 25.10.4 box:
disk.query, core.get_jobs, system.info, disk.temperatures, core.job_abort,
disk.wipe. The smart.test.* namespace was removed in 25.10; those two methods
are the SSH-superseded fallback and now raise cleanly (caller already skips).

disk.wipe is a @job method: over JSON-RPC a job call returns the integer job id
immediately (verified via pool.import_find -> core.get_jobs), same as the old
REST endpoint. Documented that contract and guard it — an untracked destructive
wipe is worse than a loud failure.

Reads keep the exponential-backoff retry; writes fail loud rather than risk a
double-execute on a dropped socket.

Concurrency: each socket owns its own pending-waiter map, so a superseded reader
running its cleanup after a reconnect can only fail its own in-flight calls and
never clobbers the current connection's pending requests.

Also:

  • mock-truenas: add a /api/current JSON-RPC WS endpoint over the same in-memory
    state so dev-against-mock keeps working; disk.wipe creates a trackable job so
    get_job() resolves. REST routes stay for /health + /debug.
  • requirements.in: promote websockets to a direct dep (already pinned at 16.0
    transitively via uvicorn[standard], so no lockfile regen).
  • tests: offline checks of the reader demux / error mapping / concurrency, the
    wipe_disk job-id contract, and stale-reader isolation (fake socket, no network).

Verified: full suite 158/158 · live read-only smoke test of the new client
against TrueNAS 25.10.4 (disks, temps, jobs, system.info) succeeded.

Co-Authored-By: Claude Opus 4.8 (1M context) noreply@anthropic.com

TrueNAS removes the /api/v2.0 REST API in 26.04. REST was stateless, so the 12s poll loop re-authenticated on every request (~130k auths/day, tripping the deprecation alarm). Switch app/truenas.py to JSON-RPC 2.0 over a single persistent, authenticated wss://.../api/current connection: requests are multiplexed by JSON-RPC id and demuxed by one background reader task, so auth happens once per connection instead of per request. Method mappings verified against a live TrueNAS SCALE 25.10.4 box: disk.query, core.get_jobs, system.info, disk.temperatures, core.job_abort, disk.wipe. The smart.test.* namespace was removed in 25.10; those two methods are the SSH-superseded fallback and now raise cleanly (caller already skips). disk.wipe is a @job method: over JSON-RPC a job call returns the integer job id immediately (verified via pool.import_find -> core.get_jobs), same as the old REST endpoint. Documented that contract and guard it — an untracked destructive wipe is worse than a loud failure. Reads keep the exponential-backoff retry; writes fail loud rather than risk a double-execute on a dropped socket. Concurrency: each socket owns its own pending-waiter map, so a superseded reader running its cleanup after a reconnect can only fail its own in-flight calls and never clobbers the current connection's pending requests. Also: - mock-truenas: add a /api/current JSON-RPC WS endpoint over the same in-memory state so dev-against-mock keeps working; disk.wipe creates a trackable job so get_job() resolves. REST routes stay for /health + /debug. - requirements.in: promote websockets to a direct dep (already pinned at 16.0 transitively via uvicorn[standard], so no lockfile regen). - tests: offline checks of the reader demux / error mapping / concurrency, the wipe_disk job-id contract, and stale-reader isolation (fake socket, no network). Verified: full suite 158/158 · live read-only smoke test of the new client against TrueNAS 25.10.4 (disks, temps, jobs, system.info) succeeded. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
truenas: migrate client from deprecated REST to JSON-RPC 2.0 over WebSocket
Some checks failed
Security scan / pip-audit (pull_request) Has been cancelled
Details
Security scan / bandit (pull_request) Has been cancelled
Details
Security scan / gitleaks (pull_request) Has been cancelled
Details
Security scan / mypy (pull_request) Has been cancelled
Details
6a850581f2 
TrueNAS removes the /api/v2.0 REST API in 26.04. REST was stateless, so the
12s poll loop re-authenticated on every request (~130k auths/day, tripping the
deprecation alarm). Switch app/truenas.py to JSON-RPC 2.0 over a single
persistent, authenticated wss://.../api/current connection: requests are
multiplexed by JSON-RPC id and demuxed by one background reader task, so auth
happens once per connection instead of per request.

Method mappings verified against a live TrueNAS SCALE 25.10.4 box:
  disk.query, core.get_jobs, system.info, disk.temperatures, core.job_abort,
  disk.wipe. The smart.test.* namespace was removed in 25.10; those two methods
  are the SSH-superseded fallback and now raise cleanly (caller already skips).

disk.wipe is a @job method: over JSON-RPC a job call returns the integer job id
immediately (verified via pool.import_find -> core.get_jobs), same as the old
REST endpoint. Documented that contract and guard it — an untracked destructive
wipe is worse than a loud failure.

Reads keep the exponential-backoff retry; writes fail loud rather than risk a
double-execute on a dropped socket.

Concurrency: each socket owns its own pending-waiter map, so a superseded reader
running its cleanup after a reconnect can only fail its own in-flight calls and
never clobbers the current connection's pending requests.

Also:
- mock-truenas: add a /api/current JSON-RPC WS endpoint over the same in-memory
  state so dev-against-mock keeps working; disk.wipe creates a trackable job so
  get_job() resolves. REST routes stay for /health + /debug.
- requirements.in: promote websockets to a direct dep (already pinned at 16.0
  transitively via uvicorn[standard], so no lockfile regen).
- tests: offline checks of the reader demux / error mapping / concurrency, the
  wipe_disk job-id contract, and stale-reader isolation (fake socket, no network).

Verified: full suite 158/158 · live read-only smoke test of the new client
against TrueNAS 25.10.4 (disks, temps, jobs, system.info) succeeded.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
brandon/nas-burnin!1
No description provided.