[gjkim42]

What type of PR is this?

/kind feature

What this PR does / why we need it:

Makes a Session a recoverable remote-control surface for one provider conversation.

• backs each Session with a one-replica StatefulSet and a headless governing Service
• adds optional Session.spec.volumeClaimTemplate; configured Sessions persist the workspace and runtime journal in a StatefulSet volume claim, while omitted storage uses emptyDir
• restores persisted conversation events after Pod replacement and marks unfinished work as interrupted without replaying its prompt
• reconnects terminal and web clients to replacement Pods while preserving accepted-request semantics
• covers StatefulSet ownership, optional storage, journaling, request delivery, and recovery behavior in unit, integration, and end-to-end tests

Which issue(s) this PR is related to:

N/A

Special notes for your reviewer:

The headless Service is only the StatefulSet governing Service. Terminal and web clients continue to use the shared Session control path rather than connecting through a per-Session Service.

Pod replacement preserves workspace and conversation state only when spec.volumeClaimTemplate is configured. A Session that omits it receives a fresh emptyDir after replacement.

Validated locally with make update, make build, make test, make verify, and make test-integration (123 controller/CLI integration specs and 12 install-integration specs). The full e2e suite runs in CI.

Does this PR introduce a user-facing change?

Sessions now run as StatefulSets and can preserve their workspace and conversation events across Pod replacement when configured with `spec.volumeClaimTemplate`; terminal and web clients reconnect automatically.

[cubic-dev-ai]

1 issue found across 27 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="internal/controller/session_controller.go">

<violation number="1" location="internal/controller/session_controller.go:574">
P2: Sessions using a workspace-mounted `extraInitContainer` rerun that container after recovery, so migrations, resets, or seed scripts can overwrite preserved workspace state. Extend recovery guarding to supported user workspace setup or reject/document non-idempotent workspace init containers for Sessions.</violation>
</file>

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[cubic-dev-ai]

P2: Sessions using a workspace-mounted extraInitContainer rerun that container after recovery, so migrations, resets, or seed scripts can overwrite preserved workspace state. Extend recovery guarding to supported user workspace setup or reject/document non-idempotent workspace init containers for Sessions.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At internal/controller/session_controller.go, line 574:

<comment>Sessions using a workspace-mounted `extraInitContainer` rerun that container after recovery, so migrations, resets, or seed scripts can overwrite preserved workspace state. Extend recovery guarding to supported user workspace setup or reject/document non-idempotent workspace init containers for Sessions.</comment>

<file context>
@@ -476,6 +563,17 @@ func (r *SessionReconciler) buildSessionPod(session *kelos.Session, workspace *k
+			break
+		}
+	}
+	if err := prepareSessionWorkspaceInit(podSpec.InitContainers); err != nil {
+		return nil, nil, err
+	}
</file context>

[knechtionscoding]

@gjkim42 this seems like an exceptional amount of code/work to handle what could be handled by having session pods use a stateful set?

[cubic-dev-ai]

1 issue found across 12 files (changes from recent commits).

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="internal/sessionruntime/server.go">

<violation number="1" location="internal/sessionruntime/server.go:306">
P1: When `journal.Append` fails in `submitMessage`, the turn was already enqueued on `s.turns` but `turn.accepted` is never closed. The `runTurns` goroutine then blocks forever on `<-turn.accepted`, preventing any further turns from being processed. This deadlocks the session runtime until the serve context is cancelled indirectly through journal failure detection.

Suggest closing `turn.accepted` before returning the error, or better, journal the event *before* enqueuing the turn so the turn is never sent unless the event is durably recorded.</violation>
</file>

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[cubic-dev-ai]

P1: When journal.Append fails in submitMessage, the turn was already enqueued on s.turns but turn.accepted is never closed. The runTurns goroutine then blocks forever on <-turn.accepted, preventing any further turns from being processed. This deadlocks the session runtime until the serve context is cancelled indirectly through journal failure detection.

Suggest closing turn.accepted before returning the error, or better, journal the event before enqueuing the turn so the turn is never sent unless the event is durably recorded.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At internal/sessionruntime/server.go, line 306:

<comment>When `journal.Append` fails in `submitMessage`, the turn was already enqueued on `s.turns` but `turn.accepted` is never closed. The `runTurns` goroutine then blocks forever on `<-turn.accepted`, preventing any further turns from being processed. This deadlocks the session runtime until the serve context is cancelled indirectly through journal failure detection.

Suggest closing `turn.accepted` before returning the error, or better, journal the event *before* enqueuing the turn so the turn is never sent unless the event is durably recorded.</comment>

<file context>
@@ -295,7 +303,9 @@ func (s *Server) submitMessage(text string) error {
 	select {
 	case s.turns <- turn:
-		s.journal.Append(Event{Type: EventUserMessage, TurnID: turn.id, Text: turn.text})
+		if err := s.journal.Append(Event{Type: EventUserMessage, RequestID: requestID, TurnID: turn.id, Text: turn.text}); err != nil {
+			return fmt.Errorf("recording Session message: %w", err)
+		}
</file context>