Add COMMAND arg for ssh commands

CHANGELOG.md

@@ -14,6 +14,9 @@ working directory. Please see the usage guide for more information about
 this command.
 
 ### Changed
+- Added optional `COMMAND` argument to `pool ssh` and `fs cluster ssh`
+commands. If `COMMAND` is specified, the command is run non-interactively
+with SSH on the target node.
 - Added some additional sanity checks in the node prep script
 - Updated TensorFlow-CPU and TensorFlow-GPU recipes to 1.1.0. Removed
 specialized Docker build for TensorFlow-GPU. Added `jobs-tb.json` files

convoy/fleet.py

@@ -1583,10 +1583,10 @@ def action_fs_cluster_status(
 
 def action_fs_cluster_ssh(
         compute_client, network_client, config, storage_cluster_id,
-        cardinal, hostname):
+        cardinal, hostname, command):
     # type: (azure.mgmt.compute.ComputeManagementClient,
     #        azure.mgmt.network.NetworkManagementClient, dict, str, int,
-    #        str) -> None
+    #        str, tuple) -> None
     """Action: Fs Cluster Ssh
     :param azure.mgmt.compute.ComputeManagementClient compute_client:
         compute client
@@ -1596,6 +1596,7 @@ def action_fs_cluster_ssh(
     :param str storage_cluster_id: storage cluster id
     :param int cardinal: cardinal number
     :param str hostname: hostname
+    :param tuple command: command
     """
     if cardinal is not None and hostname is not None:
         raise ValueError('cannot specify both cardinal and hostname options')
@@ -1608,7 +1609,7 @@ def action_fs_cluster_ssh(
             raise ValueError('invalid cardinal option value')
     remotefs.ssh_storage_cluster(
         compute_client, network_client, config, storage_cluster_id,
-        cardinal, hostname)
+        cardinal, hostname, command)
 
 
 def action_keyvault_add(keyvault_client, config, keyvault_uri, name):
@@ -1923,14 +1924,15 @@ def action_pool_dsu(batch_client, config):
     batch.del_ssh_user(batch_client, config)
 
 
-def action_pool_ssh(batch_client, config, cardinal, nodeid):
-    # type: (batchsc.BatchServiceClient, dict, int, str) -> None
+def action_pool_ssh(batch_client, config, cardinal, nodeid, command):
+    # type: (batchsc.BatchServiceClient, dict, int, str, tuple) -> None
     """Action: Pool Ssh
     :param azure.batch.batch_service_client.BatchServiceClient batch_client:
         batch client
     :param dict config: configuration dict
     :param int cardinal: cardinal node num
     :param str nodeid: node id
+    :param tuple command: command to execute
     """
     if cardinal is not None and nodeid is not None:
         raise ValueError('cannot specify both cardinal and nodeid options')
@@ -1953,11 +1955,13 @@ def action_pool_ssh(batch_client, config, cardinal, nodeid):
         batch_client, config, cardinal, nodeid)
     logger.info('connecting to node {}:{} with key {}'.format(
         ip, port, ssh_private_key))
-    util.subprocess_with_output(
-        ['ssh', '-o', 'StrictHostKeyChecking=no', '-o',
-         'UserKnownHostsFile={}'.format(os.devnull),
-         '-i', str(ssh_private_key), '-p', str(port),
-         '{}@{}'.format(pool.ssh.username, ip)])
+    ssh_cmd = ['ssh', '-o', 'StrictHostKeyChecking=no', '-o',
+               'UserKnownHostsFile={}'.format(os.devnull),
+               '-i', str(ssh_private_key), '-p', str(port),
+               '{}@{}'.format(pool.ssh.username, ip)]
+    if util.is_not_empty(command):
+        ssh_cmd.extend(command)
+    util.subprocess_with_output(ssh_cmd)
 
 
 def action_pool_delnode(batch_client, config, nodeid):

convoy/misc.py

@@ -214,8 +214,8 @@ def tunnel_tensorboard(batch_client, config, jobid, taskid, logdir, image):
              'seconds.'
              '\n\n>> Terminate your session with CTRL+C'
              '\n\n>> If you cannot terminate your session cleanly, run:'
-             '\n     shipyard pool ssh --nodeid {}'
-             '\n     sudo docker kill {}\n').format(
+             '\n     shipyard pool ssh --nodeid {} '
+             'sudo docker kill {}\n').format(
                  tb[2], task.node_info.node_id, name))
         tb_proc.wait()
     finally:

convoy/remotefs.py

@@ -2253,10 +2253,11 @@ def _get_ssh_info(
 
 
 def ssh_storage_cluster(
-        compute_client, network_client, config, sc_id, cardinal, hostname):
+        compute_client, network_client, config, sc_id, cardinal, hostname,
+        command):
     # type: (azure.mgmt.compute.ComputeManagementClient,
     #        azure.mgmt.network.NetworkManagementClient, dict, str, int,
-    #        str) -> None
+    #        str, tuple) -> None
     """SSH to a node in storage cluster
     :param azure.mgmt.compute.ComputeManagementClient compute_client:
         compute client
@@ -2266,6 +2267,7 @@ def ssh_storage_cluster(
     :param str sc_id: storage cluster id
     :param int cardinal: cardinal number
     :param str hostname: hostname
+    :param tuple command: command to execute
     """
     ssh_priv_key, port, username, ip = _get_ssh_info(
         compute_client, network_client, config, sc_id, cardinal, hostname)
@@ -2278,8 +2280,10 @@ def ssh_storage_cluster(
     logger.info(
         ('connecting to storage cluster {} virtual machine {}:{} with '
          'key {}').format(sc_id, ip, port, ssh_priv_key))
-    util.subprocess_with_output(
-        ['ssh', '-o', 'StrictHostKeyChecking=no',
-         '-o', 'UserKnownHostsFile={}'.format(os.devnull),
-         '-i', str(ssh_priv_key), '-p', str(port),
-         '{}@{}'.format(username, ip)])
+    ssh_cmd = ['ssh', '-o', 'StrictHostKeyChecking=no',
+               '-o', 'UserKnownHostsFile={}'.format(os.devnull),
+               '-i', str(ssh_priv_key), '-p', str(port),
+               '{}@{}'.format(username, ip)]
+    if util.is_not_empty(command):
+        ssh_cmd.extend(command)
+    util.subprocess_with_output(ssh_cmd)

docs/20-batch-shipyard-usage.md

@@ -286,6 +286,9 @@ specified in the configuration. This is an experimental feature.
 * `ssh` will interactively log into a virtual machine in the storage cluster.
 If neither `--cardinal` or `--hostname` are specified, `--cardinal 0` is
 assumed.
+  * `COMMAND` is an optional argument to specify the command to run. If your
+    command has switches, preface `COMMAND` with double dash as per POSIX
+    convention, e.g., `fs cluster ssh mycluster -- df -h`.
   * `--cardinal` is the zero-based cardinal number of the virtual machine in
     the storage cluster to connect to.
   * `--hostname` is the hostname of the virtual machine in the storage cluster
@@ -456,6 +459,9 @@ configuration file
   * `--wait` will wait for resize to complete
 * `ssh` will interactively log into a compute node via SSH. If neither
 `--cardinal` or `--nodeid` are specified, `--cardinal 0` is assumed.
+  * `COMMAND` is an optional argument to specify the command to run. If your
+    command has switches, preface `COMMAND` with double dash as per POSIX
+    convention, e.g., `pool ssh -- sudo docker ps -a`.
   * `--cardinal` is the zero-based cardinal number of the compute node in
     the pool to connect to as listed by `grls`
   * `--nodeid` is the node id to connect to in the pool

recipes/TensorFlow-CPU/README.md

@@ -59,8 +59,7 @@ Which will output some text similar to the following:
 >> Terminate your session with CTRL+C
 
 >> If you cannot terminate your session cleanly, run:
-     shipyard pool ssh --nodeid tvm-1518333292_4-20170428t151941z
-     sudo docker kill 9e7879b8
+     shipyard pool ssh --nodeid tvm-1518333292_4-20170428t151941z sudo docker kill 9e7879b8
 ```
 
 With a web browser, navigate to http://localhost:6006/ where Tensorboard

recipes/TensorFlow-GPU/README.md

@@ -68,8 +68,7 @@ Which will output some text similar to the following:
 >> Terminate your session with CTRL+C
 
 >> If you cannot terminate your session cleanly, run:
-     shipyard pool ssh --nodeid tvm-1518333292_4-20170428t151941z
-     sudo docker kill 9e7879b8
+     shipyard pool ssh --nodeid tvm-1518333292_4-20170428t151941z sudo docker kill 9e7879b8
 ```
 
 With a web browser, navigate to http://localhost:6006/ where Tensorboard

shipyard.py

@@ -822,17 +822,18 @@ def fs_cluster_status(ctx, storage_cluster_id, detail, hosts):
     type=int)
 @click.option(
     '--hostname', help='Hostname of remote fs vm to connect to')
+@click.argument('command', nargs=-1)
 @common_options
 @fs_cluster_options
 @aad_options
 @pass_cli_context
-def fs_cluster_ssh(ctx, storage_cluster_id, cardinal, hostname):
+def fs_cluster_ssh(ctx, storage_cluster_id, cardinal, hostname, command):
     """Interactively login via SSH to a filesystem storage cluster virtual
     machine in Azure"""
     ctx.initialize_for_fs()
     convoy.fleet.action_fs_cluster_ssh(
         ctx.compute_client, ctx.network_client, ctx.config,
-        storage_cluster_id, cardinal, hostname)
+        storage_cluster_id, cardinal, hostname, command)
 
 
 @fs.group()
@@ -1157,16 +1158,17 @@ def pool_dsu(ctx):
     type=int)
 @click.option(
     '--nodeid', help='NodeId of compute node in pool to connect to')
+@click.argument('command', nargs=-1)
 @common_options
 @batch_options
 @keyvault_options
 @aad_options
 @pass_cli_context
-def pool_ssh(ctx, cardinal, nodeid):
+def pool_ssh(ctx, cardinal, nodeid, command):
     """Interactively login via SSH to a node in the pool"""
     ctx.initialize_for_batch()
     convoy.fleet.action_pool_ssh(
-        ctx.batch_client, ctx.config, cardinal, nodeid)
+        ctx.batch_client, ctx.config, cardinal, nodeid, command)
 
 
 @pool.command('delnode')