fix(sync): stop watchdog redundant resyncs on healthy event stream (#1829)

The stale-event check excluded heartbeats from lastActiveEventAt, so a
quiet-but-connected session (only receiving heartbeats) tripped the 20s
stale timer and triggered a full resync every ~15s. This re-fetched
listPendingQuestions, listPendingPermissions, session.get, and
session.messages despite the event stream being healthy.

Track all stream activity (including heartbeats) in a global
lastStreamActivityAt ref. The stale check now only fires when no events
at all arrive for 20s, meaning the stream is genuinely dead.

Resyncs still fire correctly on genuine reconnects, transport switches,
and status-poll escalation when a real discrepancy is detected.

Fixes #1656
This commit is contained in:
Tom Rochette
2026-06-29 00:23:27 +03:00
committed by GitHub
parent 7ee893e1e5
commit 8f2e058b27
2 changed files with 83 additions and 19 deletions
+30 -19
View File
@@ -418,11 +418,6 @@ function toSessionStatus(status: Awaited<ReturnType<typeof opencodeClient.getSes
return undefined
}
function isStreamHeartbeatEvent(payload: Event): boolean {
const type = (payload as { type?: unknown }).type
return type === "server.heartbeat" || type === "openchamber:heartbeat"
}
function getActiveSessionCandidateIds(directory: string, state: DirectoryStore): string[] {
return getReconnectCandidateSessionIds(state, {
directory,
@@ -522,6 +517,27 @@ export function needsSnapshotAfterStatusPoll(
return Boolean(currentStatus && currentStatus.type !== "idle")
}
// Decide whether the event stream is genuinely stale and warrants a full
// resync. Uses stream activity that includes heartbeats, so a quiet-but-
// connected session (only receiving heartbeats) is NOT considered stale.
// A stale signal means no events at all — including no heartbeats — for the
// configured threshold, which is strong evidence the connection is dead.
// Returns false when lastStreamActivityAt is 0 (no events received yet),
// so the watchdog does not fire before the stream has delivered its first
// heartbeat.
export function shouldTriggerStaleResync(
lastStreamActivityAt: number,
lastFullResyncAt: number,
now: number,
staleThresholdMs: number = ACTIVE_SESSION_STALE_EVENT_MS,
resyncCooldownMs: number = ACTIVE_SESSION_FULL_RESYNC_COOLDOWN_MS,
): boolean {
if (lastStreamActivityAt <= 0) return false
if (now - lastStreamActivityAt < staleThresholdMs) return false
if (now - lastFullResyncAt < resyncCooldownMs) return false
return true
}
type EventRoutingIndex = {
sessionDirectoryById: Map<string, string>
messageSessionById: Map<string, string>
@@ -1575,7 +1591,7 @@ export function SyncProvider(props: {
const routingIndexRef = useRef<EventRoutingIndex | null>(null)
if (!routingIndexRef.current) routingIndexRef.current = createEventRoutingIndex()
const routingIndex = routingIndexRef.current
const lastActiveEventAtByDirectoryRef = useRef(new Map<string, number>())
const lastStreamActivityAtRef = useRef(0)
const lastStatusPollAtByDirectoryRef = useRef(new Map<string, number>())
const lastFullResyncAtByDirectoryRef = useRef(new Map<string, number>())
const lastChildDiscoveryAtByDirectoryRef = useRef(new Map<string, number>())
@@ -1759,9 +1775,13 @@ export function SyncProvider(props: {
return resolveDirectoryFromRoutingIndex(routingIndex, directory, payload, childStores)
},
onEvent: (directory, payload) => {
if (!isStreamHeartbeatEvent(payload)) {
lastActiveEventAtByDirectoryRef.current.set(directory, Date.now())
}
// Track ALL stream activity (including heartbeats) as proof of
// connection health. The watchdog stale check uses this to distinguish
// a genuinely dead stream (no heartbeats for 20s) from a quiet-but-
// connected session that is only receiving heartbeats. Excluding
// heartbeats here caused issue #1656: the stale timer fired for any
// quiet session, triggering redundant full resyncs every ~15s.
lastStreamActivityAtRef.current = Date.now()
dispatchVSCodeRuntimeNotificationEvent(directory, payload)
if (payload.type === "installation.update-available") {
const version = typeof (payload.properties as { version?: unknown })?.version === "string"
@@ -1901,28 +1921,19 @@ export function SyncProvider(props: {
const state = store.getState()
const candidateSessionIds = getActiveSessionCandidateIds(directory, state)
if (candidateSessionIds.length === 0) {
lastActiveEventAtByDirectoryRef.current.delete(directory)
lastStatusPollAtByDirectoryRef.current.delete(directory)
lastFullResyncAtByDirectoryRef.current.delete(directory)
continue
}
if (!lastActiveEventAtByDirectoryRef.current.has(directory)) {
lastActiveEventAtByDirectoryRef.current.set(directory, now)
}
const lastStatusPollAt = lastStatusPollAtByDirectoryRef.current.get(directory) ?? 0
if (now - lastStatusPollAt >= ACTIVE_SESSION_STATUS_POLL_INTERVAL_MS) {
lastStatusPollAtByDirectoryRef.current.set(directory, now)
void pollDirectoryStatuses(directory, store, candidateSessionIds).catch(() => undefined)
}
const lastActiveEventAt = lastActiveEventAtByDirectoryRef.current.get(directory) ?? now
const lastFullResyncAt = lastFullResyncAtByDirectoryRef.current.get(directory) ?? 0
if (
now - lastActiveEventAt >= ACTIVE_SESSION_STALE_EVENT_MS
&& now - lastFullResyncAt >= ACTIVE_SESSION_FULL_RESYNC_COOLDOWN_MS
) {
if (shouldTriggerStaleResync(lastStreamActivityAtRef.current, lastFullResyncAt, now)) {
pipelineReconnectRef.current?.("active_stream_stale")
triggerDirectoryResync(directory)
}