Nothing ever removed a submission: users could not retract a pending row, no
route deleted blobs or rows, and the reaper never touches uploads — so every
upload accumulated on the 5 GiB PVC forever and the only cleanup was SQL or
kubectl against the store.
- Blobs.Delete on both transports (local: RemoveAll of the id-namespaced dir,
id re-validated at the boundary; S3: idempotent object DELETE).
- Store: DeleteSubmission (admin, any status) and DeletePendingSubmission
(owner+pending CAS — a reviewed row can never be withdrawn out from under
its build).
- Manager.Delete / Manager.Withdraw delete the ROW first (under the CAS for
withdraw) and the blob after, so a live row can never point at a reaped
blob; a cleanup failure names the orphan explicitly instead of failing mute.
- API: DELETE /me/submissions/{id} (withdraw, app tier) and
DELETE /api/v1/submissions/{id} (admin) both return the row as it was;
audit events submission.withdraw / submission.delete; openapi documents both
paths; admin route pinned in the admin-only table.
- Panel: two-step withdraw on a pending row (frees the pending slot and the
storage budget); two-step delete on every admin row; zh/en copy; wire tests.
Unit: submit (withdraw happy path / wrong owner / reviewed row / no transport /
blob-cleanup failure), local+S3 delete idempotence, api handlers (200/404/409/
503 + route tier); pgint: withdraw CAS + admin delete exactly-once.
go vet/go test/gofmt clean; panel vitest 120 + typecheck green.
A logged-in user could file submissions without bound and stream a 1 GiB
context per submission. The only limits were the single-blob size cap and the
5 GiB uploads PVC (platform/workloads.go); nothing counted a user's rows or
bytes, so one account could fill the volume and every other user's upload
would start failing.
- Create: per-user pending_review cap (default 5) — the review queue cannot
be parked full of one account's rows. Check-then-insert, documented soft.
- UploadContext: per-user stored-context budget (default 2 GiB) charged
against the blob store's REAL sizes (new Blobs.Size on local/S3 stores), so
the sum cannot drift from the volume; the write is capped at the remaining
budget, so the excess is refused before it is persisted, and a re-upload is
charged only for its new bytes.
- API: per-user create/upload throttles (30s/15s, cmd/felis-wired) on a
dedicated cooldown keyspace, reserve→release so a failed attempt never
burns the window and a burst collapses to one winner; ErrQuotaExceeded →
403 submission_quota_exceeded (distinct from the 400 an oversize blob
gets), 429 submission_cooldown for the throttles.
- Panel: zh/en copy for both codes; openapi documents 403/429 on the two
user routes; pgint covers the pending-queue count.
Unit tests: submit package (cap, budget boundary/exact-fit/replacement,
oversize-vs-quota split) and api handlers (quota 403 both paths, throttle
429 + recovery + failure-release). go vet/go test/gofmt clean; panel
vitest 118 + typecheck green.
login/lobby carry reserved names, so every per-server route rejects them —
yet the cockpit offered claim/stop/wake and a console link on their rows,
each answering 400 bad_name. The fleet view now marks them (system:true,
shared naming.IsSystemServer) and the panel renders a plain label instead
of dead actions.
/me/submissions (and the admin queue) now attach build_status/build_error by
a read-only Builder.Get — until now a failed build was visible only on the
admin-tier /images/build routes, so the person who submitted the modpack
never learned the build died. A missing build row renders as "no outcome";
any other lookup failure surfaces instead of being swallowed. The panel's
My Submissions page renders the outcome in the expanded row, localised.
A submitted modpack was durable but unreadable: the uploads PVC cannot cross
namespaces (felis-api mounts it; Kaniko runs in felis-build) and the s3 lane
handed the sandboxed build Pod no credentials, so NO user build could ever
consume its context. The transport is now the API itself:
- submit: derived context refs become the internal-face URL
/api/v1/internal/submissions/{id}/context (service-token gated), and Blobs
gains Open (local + s3) with an ErrBlobNotFound sentinel for the route's 404.
- api: serves that route on the internal face only (openapi.yaml updated; the
route-coverage test enforces it).
- build: an http(s) context renders a context-fetch initContainer (the felis
image's new fetch-context entrypoint) that streams the blob with the
namespace-local service-token Secret — never mounted into Kaniko — and
extracts it under a zip-slip guard into a size-limited emptyDir that Kaniko
reads read-only as --context=/context.
- platform/install: the api Deployment carries its own internal base URL; the
build namespace gets the token Secret through the existing replica mechanism
(bootstrap.sh + felis setup); the build egress lock opens exactly the control
namespace on the internal port.
- cmd/felis: fetch-context entrypoint (registered, documented, unit-tested for
escapes/symlinks/non-gzip).
Tests cover rendering, hardening, the s3/local Open paths, and the route's
404/503 mapping. Verified next on the real single-node cluster with Kaniko.
- [registry] gains kaniko_image / trivy_image / build_cpu_limit /
build_mem_limit overrides; empty keeps the compiled-in defaults. An
air-gapped or mirrored install has no route to gcr.io/aquasec (the
build egress policy allows only DNS + registry + package mirrors), so
builds previously could not even start their executors.
- deferred-seams: the uploads-context entry now records WHY a mount is
impossible (PVCs cannot cross namespaces) and that the s3 lane also
lacks credentials in the build Pod — options captured for the real fix.
- troubleshooting 8e (executor ImagePullBackOff + the overrides),
13b rewritten (verified eviction refusal, 5m pressure-transition,
image-GC recovery), 15 (upgrade/rollback runbook for Recreate).
- Backup semantics decided and documented: a backup is the whole /data
volume (worlds + config + plugins + cache) and a restore rolls all of
it back — OpenAPI/README wording updated to match (same-tag images are
still watched for regressions by the openapi parity gate).
Backup and restore only enqueue a cluster Job; a later failure left its
only trace in that Job object, invisible without kubectl. Add
GET /api/v1/servers/{name}/jobs (owner-or-admin) projecting the newest
20 managed Jobs (felis-backup / felis-restore) as
running|succeeded|failed with message and timestamps. Nil reader -> 503
jobs_unavailable, mirroring the backup/restore feature gates. RBAC gains
jobs:list; OpenAPI parity updated.
The hasJoined contract listed only 200 and 204 and named authlib as
the caller. The handler now answers four more ways, and a proxy
operator reading the contract could not tell a refused login from a
down source.
- 204 also covers a missing or oversized parameter (no source is
asked), a third-party name that is not a legal Minecraft username,
and an identity id that does not parse.
- 400 for a request that declares a body. There is no response body,
and the connection is closed.
- 500 when the bar-list lookup fails, with the usual error body.
- 503 when no source validated and at least one failed, since that
source's player may be the one logging in.
The three query parameters now carry the 64-byte cap. The profile name
says a third-party player holding a registered Mojang name gets it
back prefixed and cut to 16 characters. The description names
Velocity, drops the "thin login hook" that does not exist, and says
that a non-200, non-204 answer makes Velocity report the auth servers
as down.
A live install passed the SMTP setup screen and then failed every one-time
code with a bare `internal error`. Four separate defects had to line up for
that, and each is fixed here.
The relay was configured with `from = noreply@<domain-A>` on an account
authenticated as `<user>@<domain-B>`. Providers that validate sender identity
— Fastmail among them — answer MAIL FROM with an unconditional 250 and only
refuse at end-of-DATA. Ping stopped at NOOP, so it never saw the refusal: the
wizard reported success, wrote the config, rolled felis-api, and every OTP
afterwards died at w.Close().
Ping now runs the same transaction a real code takes — connect, (STARTTLS,)
AUTH, MAIL FROM, RCPT TO, DATA — delivering one self-test message to the From
address, and SendOTP and Ping share deliver() so the check cannot drift from
the thing it checks. The self-test recipient cannot cause a false negative:
an authenticated submission relay accepts RCPT for any destination by
definition, while the sender identity it does validate is exactly what we
want tested. The setup screen now says a message will be sent, names the
address it went to, and warns that From must be an address the account is
allowed to send as.
A relay refusal also answered 500 `internal`, which reads as a broken panel
and sends the operator hunting through handler code instead of their [smtp]
block. It is now 502 `mail_undeliverable`, mapped inside deliverOTP so all
four doors that mail a code (onboarding, email login, op-login, migrate
step-up) answer alike. The relay's own text stays out of the response — it
can name the SMTP account, and these routes are reachable by any signed-in
player — and goes to the log instead.
writeError logged nothing when it collapsed an unmapped error to 500, so an
operator holding an `internal error` had nothing to grep for and diagnosis
degraded into guessing against a live install. It now logs the method, path,
wrapped chain and the same request_id the caller is shown.
Finally, write_felis_toml regenerated the config wholesale and never emitted
[smtp], so re-running the installer — the documented way to update felis-api —
silently erased a working relay and reverted OTP delivery to the no-Mailer
path, logging codes instead of sending them. It now carries the block forward,
cached on first read because the host toml is clobbered before the pod toml is
written. Same defect family as the root_domain loss fixed in ecbeb20: a
generated file holding a hand-set value with no carry-forward.
Tests cover the case a MAIL FROM probe cannot see: a fake relay that answers
250 to MAIL FROM and 550 at end-of-DATA must fail both Ping and SendOTP, and
the 502 must carry a distinct machine code without leaking the relay's text.
Give an owner a way to repair the one failure no other endpoint covers: a
server that will not boot because a single line of server.properties or a
plugin's YAML is wrong. Until now that needed a human with cluster access.
felis-api cannot touch a world in-process — the world PVC is ReadWriteOnce
and its lifecycle belongs to the operator's StatefulSet — so the work runs
as a one-shot Job, and the server must be stopped first because a running
one holds the volume. That is the same constraint that shapes restore and
backup, and the handlers enforce the stopped gate the same way.
What is different is that the caller wants the OUTPUT, not just the side
effect. The Job prints its result to stdout and felis-api reads it back
through the pods/log subresource, which needs no permission felis-api does
not already hold: jobs:create, pods:list, pods/log:get. No pods/exec, no
pods/portforward, not even pods:get. The price is latency — every operation
is a Pod schedule — which is why this is a repair tool and not a file
manager.
Containment is structural, not textual. Every filesystem access goes through
os.Root, the stdlib's escape-proof directory handle, which resolves each
component against the open root descriptor and refuses "..", absolute paths,
and symlinks leading outside. The string-prefix check used elsewhere is not
reused here: it validates a path as text and then opens it as a path, and a
world directory holds attacker-influenced content, so a symlink swapped in
between those two steps is a live threat rather than a theoretical one.
os.Root has no such window because the check and the open are one operation.
The Job's isolation is a strict subset of a restore Pod's: the weak
felis-restore SA with its token auto-mount disabled, exactly one volume (the
world PVC, mounted read-only for list and read so two of the three
operations cannot mutate anything), no Secret, no ConfigMap, no database
URL, non-root with an fsGroup matching the operator's so a written file is
readable by the server that later mounts it, and backoffLimit 0 so a failed
write is never silently retried as a second write.
Two limits on the surface are worth stating plainly, because the mount is
the server's whole working directory rather than a config subtree:
* A write accepts arbitrary bytes at any path, so an owner can place a
loadable plugin jar. This is deliberate — it is what a hosting panel's
file manager does, scoped to a server the caller already owns and
already drives through /command — but it is the one owner-tier route
that lands executable code in a backend pod, since images are
admin-only and modpack submissions need an admin verdict.
* config/paper-global.yml is refused on read. felis-lobby's entrypoint
writes FELIS_FORWARDING_SECRET into it on every boot, and that value is
identical on every backend, so reading it from a server you own would
hand you the handshake key for everyone else's. It is the only path in
the mount that is not the caller's own data, and therefore the only
denial. The comparison is on the cleaned path, or ./config/... would
walk straight through it.
Writing that file is still allowed: it leaks nothing, and the entrypoint
rewrites it whole on every boot regardless.
The write body's content field is a *[]byte rather than a []byte for the
reason permissionRequest.Value is a *bool — a plain slice makes absent,
null, and empty indistinguishable, so a body of {} would decode to nil and
truncate the target to zero bytes while answering 200, destroying the very
config the caller opened the editor to repair.
Lemon-miaow
flyemoji