Command: alloc exec
alloc exec command runs a command in a running allocation.
nomad alloc exec [options] <allocation> <command> [<args>...]
The nomad exec command can be use to run commands inside a running task/allocation.
Use cases are for inspecting container state, debugging a failed application without needing ssh access into the node that's running the allocation.
This command executes the command in the given task in the allocation. If the
allocation is only running a single task, the task name can be omitted.
-job option may be used in which case a random allocation from
the given job will be chosen.
When ACLs are enabled, this command requires a token with the
list-jobs capabilities for the allocation's namespace. If
the task driver does not have file system isolation (as with
this command requires the
capabilities for the allocation's namespace.
-address=<addr>: The address of the Nomad server. Overrides the
NOMAD_ADDRenvironment variable if set. Defaults to
-region=<region>: The region of the Nomad server to forward commands to. Overrides the
NOMAD_REGIONenvironment variable if set. Defaults to the Agent's local region.
-namespace=<namespace>: The target namespace for queries and actions bound to a namespace. Overrides the
NOMAD_NAMESPACEenvironment variable if set. If set to
'*', subcommands which support this functionality query all namespaces authorized to user. Defaults to the "default" namespace.
-no-color: Disables colored command output. Alternatively,
NOMAD_CLI_NO_COLORmay be set. This option takes precedence over
-force-color: Forces colored command output. This can be used in cases where the usual terminal detection fails. Alternatively,
NOMAD_CLI_FORCE_COLORmay be set. This option has no effect if
-no-coloris also used.
-ca-cert=<path>: Path to a PEM encoded CA cert file to use to verify the Nomad server SSL certificate. Overrides the
NOMAD_CACERTenvironment variable if set.
-ca-path=<path>: Path to a directory of PEM encoded CA cert files to verify the Nomad server SSL certificate. If both
-ca-certis used. Overrides the
NOMAD_CAPATHenvironment variable if set.
-client-cert=<path>: Path to a PEM encoded client certificate for TLS authentication to the Nomad server. Must also specify
-client-key. Overrides the
NOMAD_CLIENT_CERTenvironment variable if set.
-client-key=<path>: Path to an unencrypted PEM encoded private key matching the client certificate from
-client-cert. Overrides the
NOMAD_CLIENT_KEYenvironment variable if set.
-tls-server-name=<value>: The server name to use as the SNI host when connecting via TLS. Overrides the
NOMAD_TLS_SERVER_NAMEenvironment variable if set.
-tls-skip-verify: Do not verify TLS certificate. This is highly not recommended. Verification will also be skipped if
-token: The SecretID of an ACL token to use to authenticate API requests with. Overrides the
NOMAD_TOKENenvironment variable if set.
-task: Sets the task to exec command in.
-job: Use a random allocation from the specified job or job ID prefix, preferring a running allocation.
-i: Pass stdin to the container, defaults to true. Pass
-i=falseto disable explicitly.
-t: Allocate a pseudo-tty, defaults to true if stdin is detected to be a tty session. Pass
-t=falseto disable explicitly.
-e<escape_char>: Sets the escape character for sessions with a pty (default: '~'). The escape character is only recognized at the beginning of a line. The escape character followed by a dot ('.') closes the connection. Setting the character to 'none' disables any escapes and makes the session fully transparent.
To start an interactive debugging session in a particular alloc, invoke exec command with your desired shell available inside the task:
$ nomad alloc exec eb17e557 /bin/bash root@eb17e557:/data# # now run any debugging commands inside container root@eb17e557:/data# # ps -ef
To run a command and stream results without starting an interactive shell, you can pass the command and its arguments to exec directly:
$ nomad alloc exec eb17e557 cat /etc/resolv.conf ...
When passing command arguments to be evaluated in task, you may need to ensure
that your host shell doesn't interpolate values before invoking
For example, the following command would return the environment variable on
operator shell rather than task containers:
$ nomad alloc exec eb17e557 echo $NOMAD_ALLOC_ID # wrong ...
Here, we must start a shell in task to interpolate
$NOMAD_ALLOC_ID, and quote
command or use the heredoc syntax
$ nomad alloc exec eb17e557 /bin/sh -c 'echo $NOMAD_ALLOC_ID' eb17e557-443e-4c51-c049-5bba7a9850bc $ # by using heredoc and passing command in stdin $ nomad alloc exec eb17e557 /bin/sh <<'EOF' > echo $NOMAD_ALLOC_ID > EOF eb17e557-443e-4c51-c049-5bba7a9850bc
This technique applies when aiming to run a shell pipeline without streaming intermediate command output across the network:
$ nomad alloc exec eb17e557 /bin/sh -c 'cat /output | sort | uniq -c | sort -rn | head -n 5' ...
-job flag causes a random allocation of the specified job to be
nomad alloc exec -job <job-id> <command> [<args>...]
Choosing a specific allocation is useful for debugging issues with a specific
instance of a service. For other operations using the
-job flag may be more
convenient than looking up an allocation ID to use.
alloc exec is enabled by default to aid with debugging. Operators can disable
the feature by setting
disable_remote_exec client config
option on all clients, or a subset of clients that
run sensitive workloads.
When trying to
alloc exec for a job that has more than one task associated
with it, you may want to target a specific task.
# open a shell session in one of your allocation's tasks $ nomad alloc exec -i -t -task mytask a1827f93 /bin/bash a1827f93$