Remove host_info.json on instance deploy if exists (#1479)

Ensures that shim will create a fresh file at the first start

Fixes: #1465

src/dstack/_internal/core/backends/remote/provisioning.py

@@ -23,6 +23,8 @@
 
 DSTACK_SHIM_ENV_FILE = "dstack-shim.env"
 
+HOST_INFO_FILE = "host_info.json"
+
 
 def sftp_upload(client: paramiko.SSHClient, path: str, body: str) -> None:
     try:
@@ -139,22 +141,35 @@ def check_dstack_shim_service(client: paramiko.SSHClient):
             raise ProvisioningError(f"The dstack-shim service doesn't start: {line.strip()}")
 
 
+def remove_host_info_if_exists(client: paramiko.SSHClient, working_dir: str) -> None:
+    file_path = f"{working_dir}/{HOST_INFO_FILE}"
+    try:
+        _, _, stderr = client.exec_command(
+            f"sudo test -e {file_path} && sudo rm {file_path}", timeout=10
+        )
+        err = stderr.read().decode().strip()
+        if err:
+            logger.debug(f"{HOST_INFO_FILE} hasn't been removed: %s", err)
+    except (paramiko.SSHException, OSError) as e:
+        raise ProvisioningError(f"remove_host_info_if_exists failed: {e}")
+
+
 def get_host_info(client: paramiko.SSHClient, working_dir: str) -> Dict[str, Any]:
     # wait host_info
     retries = 60
     iter_delay = 3
     for _ in range(retries):
         try:
             _, stdout, stderr = client.exec_command(
-                f"sudo cat {working_dir}/host_info.json", timeout=10
+                f"sudo cat {working_dir}/{HOST_INFO_FILE}", timeout=10
             )
             err = stderr.read().decode().strip()
             if err:
                 logger.debug("Retry after error: %s", err)
                 time.sleep(iter_delay)
                 continue
         except (paramiko.SSHException, OSError) as e:
-            logger.debug("Cannot run `cat host_info.json` in the remote instance: %s", e)
+            logger.debug(f"Cannot run `cat {HOST_INFO_FILE}` in the remote instance: %s", e)
         else:
             try:
                 host_info_json = stdout.read()

src/dstack/_internal/server/background/tasks/process_instances.py

@@ -23,6 +23,7 @@
     get_paramiko_connection,
     get_shim_healthcheck,
     host_info_to_instance_type,
+    remove_host_info_if_exists,
     run_pre_start_commands,
     run_shim_as_systemd_service,
     upload_envs,
@@ -169,6 +170,9 @@ def deploy_instance(
         upload_envs(client, DSTACK_WORKING_DIR, shim_envs)
         logger.debug("The dstack-shim environment variables have been installed")
 
+        # Ensure host info file does not exist
+        remove_host_info_if_exists(client, DSTACK_WORKING_DIR)
+
         # Run dstack-shim as a systemd service
         run_shim_as_systemd_service(
             client=client,