ray.init bug on windows linux subsystem

[robertnishihara]

The following happens when doing ray.init on WSL for Windows. This can be fixed by doing pip install psutil.

CalledProcessError                        Traceback (most recent call last)
<ipython-input-2-193b426d1a4d> in <module>()
      2 import ray
      3 #ray.init("localhost:6380")
----> 4 ray.init()


~/miniconda3/envs/tf-latest/lib/python3.6/site-packages/ray/worker.py in init(redis_address, num_cpus, num_gpus, resources, object_store_memory, node_ip_address, object_id_seed, num_workers, local_mode, driver_mode, redirect_worker_output, redirect_output, ignore_reinit_error, num_redis_shards, redis_max_clients, redis_protected_mode, plasma_directory, huge_pages, include_webui, use_raylet, configure_logging, logging_level, logging_format)

   1864         include_webui=include_webui,
   1865         object_store_memory=object_store_memory,
-> 1866         use_raylet=use_raylet)
   1867     for hook in _post_init_hooks:
   1868         hook()

 
~/miniconda3/envs/tf-latest/lib/python3.6/site-packages/ray/worker.py in _init(address_info, start_ray_local, object_id_seed, num_workers, num_local_schedulers, object_store_memory, local_mode, driver_mode, redirect_worker_output, redirect_output, start_workers_from_local_scheduler, num_cpus, num_gpus, resources, num_redis_shards, redis_max_clients, plasma_directory, huge_pages, include_webui, use_raylet)

   1659             huge_pages=huge_pages,
   1660             include_webui=include_webui,
-> 1661             use_raylet=use_raylet)
   1662     else:
   1663         if redis_address is None:


~/miniconda3/envs/tf-latest/lib/python3.6/site-packages/ray/services.py in start_ray_head(address_info, node_ip_address, redis_port, redis_shard_ports, num_workers, num_local_schedulers, object_store_memory, worker_path, cleanup, redirect_worker_output, redirect_output, start_workers_from_local_scheduler, resources, num_redis_shards, redis_max_clients, redis_protected_mode, include_webui, plasma_directory, huge_pages, autoscaling_config, use_raylet)

   1800         huge_pages=huge_pages,
   1801         autoscaling_config=autoscaling_config,
-> 1802         use_raylet=use_raylet)
   1803
   1804
 

~/miniconda3/envs/tf-latest/lib/python3.6/site-packages/ray/services.py in start_ray_processes(address_info, node_ip_address, redis_port, redis_shard_ports, num_workers, num_local_schedulers, object_store_memory, num_redis_shards, redis_max_clients, redis_protected_mode, worker_path, cleanup, redirect_worker_output, redirect_output, include_global_scheduler, include_log_monitor, include_webui, start_workers_from_local_scheduler, resources, plasma_directory, huge_pages, autoscaling_config, use_raylet)

   1522             plasma_directory=plasma_directory,
   1523             huge_pages=huge_pages,
-> 1524             use_raylet=use_raylet)
   1525         object_store_addresses.append(object_store_address)
   1526         time.sleep(0.1)
 

~/miniconda3/envs/tf-latest/lib/python3.6/site-packages/ray/services.py in start_plasma_store(node_ip_address, redis_address, object_manager_port, store_stdout_file, store_stderr_file, manager_stdout_file, manager_stderr_file, objstore_memory, cleanup, plasma_directory, huge_pages, use_raylet)

   1122     if objstore_memory is None:
   1123         # Compute a fraction of the system memory for the Plasma store to use.
-> 1124         system_memory = ray.utils.get_system_memory()
   1125         if sys.platform == "linux" or sys.platform == "linux2":
   1126             # On linux we use /dev/shm, its size is half the size of the
 

~/miniconda3/envs/tf-latest/lib/python3.6/site-packages/ray/utils.py in get_system_memory()

    324         # Handle Linux.
    325         bytes_in_kilobyte = 1024
--> 326         return vmstat("total memory") * bytes_in_kilobyte
    327     else:
    328         # Handle MacOS.
 

~/miniconda3/envs/tf-latest/lib/python3.6/site-packages/ray/utils.py in vmstat(stat)

    279         The parsed output.
    280     """
--> 281     out = subprocess.check_output(["vmstat", "-s"])
    282     stat = stat.encode("ascii")
    283     for line in out.split(b"\n"):
 

~/miniconda3/envs/tf-latest/lib/python3.6/subprocess.py in check_output(timeout, *popenargs, **kwargs)

    334
    335     return run(*popenargs, stdout=PIPE, timeout=timeout, check=True,
--> 336                **kwargs).stdout
    337
    338
 

~/miniconda3/envs/tf-latest/lib/python3.6/subprocess.py in run(input, timeout, check, *popenargs, **kwargs)

    416         if check and retcode:
    417             raise CalledProcessError(retcode, process.args,
--> 418                                      output=stdout, stderr=stderr)
    419     return CompletedProcess(process.args, retcode, stdout, stderr)
    420

CalledProcessError: Command '['vmstat', '-s']' returned non-zero exit status 102.

The issue may be that /proc doesn't exist.

[animesh]

Thanks @robertnishihara

sudo pip3 install psutil
sudo pip3 install setproctitle

saved the day 👍

[ninafiona]

Unfortunately the solution 'pip3 install psutil setproctitle' didn't work for me.
I tried to run ray.init() and get the error 'Could not establish connection to redis ... (context.err = 1)'.
It works on my other Linux machine, but not on the WSL. Any other ideas?

[mehrdadn]

Closing this as it is now tracked on #6900.