persist restic cache and backup restore fixes

src/lib/server/backups/helpers.ts

@@ -5,6 +5,7 @@
  * no globals, no DB. Safe to call in any context and to unit-test in isolation.
  */
 
+import { join } from 'node:path';
 import { isValidCron } from '../scheduler/cron-utils';
 import { privateIpReason, dangerousHostReason, isSafeWebhookUrl, isSafeNotificationUrl } from '../url-safety';
 import { BackupError, isLocalRepo } from './models';
@@ -297,7 +298,12 @@ export function isRepoNotInitializedError(error: unknown): boolean {
 export function classifyBackupError(error: unknown): string {
 	if (isRepoNotInitializedError(error)) return 'REPO_NOT_INIT';
 	const raw = (error instanceof Error ? error.message : String(error ?? '')).toLowerCase();
-	if (raw.includes('sigterm') || raw.includes('sigkill') || raw.includes('timed out') || raw.includes('timeout')) return 'RESTIC_TIMEOUT';
+	// A killed/cancelled restic (slow remote backend, watchdog timeout) reports its
+	// OWN exit as `signal terminated` / `context canceled`, and restic then prints a
+	// MISLEADING `code:12 wrong password or no key found` because it was interrupted
+	// mid key-load. Catch the termination markers FIRST so a timeout is never
+	// mis-surfaced to the user as a bad password.
+	if (raw.includes('sigterm') || raw.includes('sigkill') || raw.includes('signal terminated') || raw.includes('signal killed') || raw.includes('context canceled') || raw.includes('context cancelled') || raw.includes('timed out') || raw.includes('timeout')) return 'RESTIC_TIMEOUT';
 	if (raw.includes('wrong password') || raw.includes('invalid password') || raw.includes('no key found')) return 'WRONG_PASSWORD';
 	if (raw.includes('already locked') || raw.includes('unable to create lock') || raw.includes('repository is already locked')) return 'REPO_LOCKED';
 	if (raw.includes('no volumes') || raw.includes('nothing to backup')) return 'NO_VOLUMES';
@@ -399,6 +405,16 @@ export function buildResticEnv(
 	for (const [k, v] of Object.entries(procEnv)) {
 		if (v !== undefined && RESTIC_BASE_ENV_ALLOW.has(k)) env[k] = v;
 	}
+	// Persist restic's cache under DATA_DIR (a durable volume) instead of the default
+	// $HOME/.cache/restic on the container's ephemeral RW layer, which is wiped on every
+	// redeploy. A cold cache re-fetches the repo index from the remote backend (backblaze/
+	// S3) on the first ls/snapshots/dump after a redeploy - ~3x slower and enough to trip a
+	// reverse proxy's idle timeout. Only for the host-side restic (this fn feeds runLocal +
+	// destination test); the helper container builds its env separately (buildHelperEnv). An
+	// operator-set RESTIC_CACHE_DIR (copied in above) wins.
+	if (!env.RESTIC_CACHE_DIR) {
+		env.RESTIC_CACHE_DIR = join(procEnv.DATA_DIR || '/app/data', '.restic-cache');
+	}
 	env.RESTIC_REPOSITORY = opts.repository;
 	env.RESTIC_PASSWORD = opts.password;
 	for (const [k, v] of Object.entries(opts.envVars || {})) {

src/lib/server/backups/index.ts

@@ -1046,7 +1046,12 @@ function reader() {
 async function listSnapshotVolumes(destination: any, snapshotId: string): Promise<string[]> {
 	const { isReservedVolumeKey } = await import('./stackdir-plan');
 	const run = await restic.runLocal(destination, ['ls', '--json', '--no-lock', snapshotId, '/volumes/']);
-	if (run.exitCode !== 0) return [];
+	// A FAILED `ls` (timeout / killed / slow remote backend) is NOT "the snapshot has no
+	// volumes" - returning [] there makes restore report a MISLEADING "not found in snapshot,
+	// Available: none" for a snapshot that actually holds data. Fail loud with the real reason.
+	if (run.exitCode !== 0) {
+		throw new BackupError('RESTIC', `could not list the snapshot's volumes: ${run.stderr.trim() || 'restic ls failed'}`, { exitCode: run.exitCode });
+	}
 	const names = new Set<string>();
 	for (const line of run.stdout.split('\n')) {
 		try {

src/routes/containers/RestoreModal.svelte

@@ -1047,8 +1047,12 @@
 						disabled={!canRun || restoring}
 						variant={mode === 'in-place' ? 'destructive' : 'default'}
 					>
-						{#if restoring}<Loader2 class="mr-1.5 h-4 w-4 animate-spin" />{:else}<Play class="mr-1.5 h-4 w-4" />{/if}
-						{mode === 'in-place' ? 'Overwrite & restore' : (postRestore !== 'none' ? 'Restore & start' : 'Restore')}
+						{#if restoring}<Loader2 class="mr-1.5 h-4 w-4 animate-spin" />{:else if targetPreviewLoading}<Loader2 class="mr-1.5 h-4 w-4 animate-spin" />{:else}<Play class="mr-1.5 h-4 w-4" />{/if}
+						{#if targetPreviewLoading && !restoring}
+							Checking target&hellip;
+						{:else}
+							{mode === 'in-place' ? 'Overwrite & restore' : (postRestore !== 'none' ? 'Restore & start' : 'Restore')}
+						{/if}
 					</Button>
 				{/if}
 			{/if}