Enhance backend management and performance tuning

- Introduced remounting logic for data directories to ensure correct bindings before restarting services, improving reliability during performance tuning.
- Added functions to sanitize and ensure absolute paths for backend scripts, preventing issues with relative paths in bind mounts.
- Enhanced the `patch_backends_extra_args` function to clean up corrupted entries and ensure proper configuration of backend parameters.
- Updated tests to validate the new sanitization and path handling functionalities, ensuring robustness in backend management.

src/gpu_rent/session.py

@@ -178,12 +178,36 @@ def _bind_access(
         clock.mark("Idle", log)
         try:
             if tune_swarm_perf(cfg, ip, log):
+                # git reset / prior boots can drop binds; remount before restart.
+                run_ssh(
+                    cfg,
+                    ip,
+                    "sudo -n bash -c '"
+                    "for p in dlbackend Data Models Output; do "
+                    "src=/mnt/swarm_data/$p; dst=/opt/swarmui/$p; "
+                    "mkdir -p \"$src\" \"$dst\"; "
+                    "cur=$(findmnt -n -o SOURCE --target \"$dst\" 2>/dev/null || true); "
+                    "if [[ \"$cur\" != \"$src\" ]]; then "
+                    "umount \"$dst\" 2>/dev/null || umount -l \"$dst\" 2>/dev/null || true; "
+                    "mount --bind \"$src\" \"$dst\" || true; "
+                    "echo remounted $dst; "
+                    "fi; "
+                    "done'",
+                    check=False,
+                    timeout=60,
+                )
                 log("systemctl restart swarmui (perf ExtraArgs)")
                 run_ssh(cfg, ip, "sudo -n systemctl restart swarmui", timeout=120)
                 try:
                     wait_backend_idle(cfg, ip, log)
                 except CloudError as exc:
-                    log(f"ready after perf: {exc}")
+                    log(f"ready after perf: {exc} — recover backends")
+                    try:
+                        ensure_swarm_comfy_installed(cfg, ip, log)
+                        wait_backend_idle(cfg, ip, log)
+                    except Exception as exc2:
+                        log(f"recover after perf: {exc2}")
+                        raise exc from exc2
             else:
                 log("perf tune: restart не нужен")
         except Exception as exc:
@@ -314,8 +338,26 @@ def cmd_up(
             state.server_id = existing.id
             state.server_name = getattr(existing, "name", None)
             if status == "ACTIVE":
-                if state.bootstrapped and state.floating_ip:
-                    # Warm re-up: skip git pull/stop cascade unless --update.
+                # Prefer FIP from state; may still need ensure below.
+                fip = state.floating_ip
+                # Warm if local state says so, OR VM still has bootstrap marker
+                # (local bootstrapped often cleared on stop while disk/VM remain).
+                vm_warm = False
+                if fip and not (state.bootstrapped and state.floating_ip):
+                    try:
+                        if probe_ssh(cfg, fip, attempts=1) == "ok":
+                            probe = run_ssh(
+                                cfg,
+                                fip,
+                                "test -f /opt/swarmui/.gpu-rent-bootstrapped && echo yes || echo no",
+                                check=False,
+                                timeout=20,
+                            ).strip()
+                            vm_warm = probe.endswith("yes") or probe == "yes"
+                    except Exception:
+                        vm_warm = False
+
+                if (state.bootstrapped and state.floating_ip) or vm_warm:
                     if update is None:
                         do_update = False
                         log(
@@ -324,6 +366,9 @@ def cmd_up(
                         )
                     else:
                         do_update = update
+                    if vm_warm and not state.bootstrapped:
+                        log("warm: маркер на VM — восстанавливаю local bootstrapped")
+                        state.bootstrapped = True
                     log("сервер уже ACTIVE — второй GPU не создаём")
                     state.phase = "bootstrapping"
                     save_state(state)
@@ -331,7 +376,6 @@ def cmd_up(
                     return state
 
                 # Bootstrap не завершён: почти всегда VM без authorized_keys.
-                fip = state.floating_ip
                 if not fip:
                     try:
                         fip, fip_id = ensure_floating_ip(