slurmrestd

Section: Slurm REST Daemon (8)
Updated: Slurm REST Daemon
Index

NAME

slurmrestd - Interface to Slurm via REST API.

SYNOPSIS

slurmrestd [OPTIONS...] <[host]:port|unix:/path/to/socket>...

DESCRIPTION

slurmrestd is REST API interface for Slurm. It can be used in two modes:

Inetd Mode: slurmrestd will read and write to STDIO. If STDIN is a socket file descriptor, then slurmrestd will detect this and use relevant functionality. If a controlling TTY is detected, interactive mode will automatically activate to provide additional logging information. This mode is designed to work with piped input, inetd, xinetd or systemd socket activation.

Listen Mode: slurmrestd will open a listening socket on each requested host:port pair or UNIX socket.

OPTIONS

[host]:port

Hostname and port to listen against. host may be an IPv4/IPv6 IP or a resolvable hostname. Hostnames are only looked up at startup and do not change for the life of the process. host is optional; if not provided, slurmrestd will listen on all network interfaces.

unix:/path/to/socket

Listen on local UNIX socket. Must have permission to create socket in filesystem.

-a <plugin>[,plugin]...

Comma-delimited list of authentication plugins to load. By default all available authentication plugins will be loaded.

list

Display a list of the possible plugins to load.

rest_auth/local

Allows authentication via UNIX sockets when auth/munge is active.
NOTE: slurmrestd and client processes must run under the same UID or the client requests will be rejected.

rest_auth/jwt

Allows authentication via TCP and UNIX sockets when AuthAltTypes=auth/jwt is active. User must specify the following HTTP cookies with each request:

X-SLURM-USER-NAME:<user name>

X-SLURM-USER-TOKEN:<JSON Web Token>

NOTE: Tokens are usually generated via calling "scontrol token".

-d <plugin>[,plugin]...

Comma-delimited list of data_parser plugins, which will determine the output format. May include optional flags denoted by '+' symbol. By default all available data_parser plugins will be loaded without any optional flags. See also -s for OpenAPI plugins controlling the available content. Defaults: all builtin supported data_parser plugins.

list[+flags]

Display a list of the possible plugins to load.

latest

Automatically replaced with latest plugin version. See relavent plugin for description and potential flags.

[data_parser/]v0.0.40[+fast]

Load data_parser/]v0.0.40 plugin to for formatting of data. Only compatible with openapi/slurmctld and openapi/slurmdbd content plugins.

+fast: Disable builtin warnings and other logic to strictly validate incoming requests. Should only ever be used in a production environment with very well tested clients and potentially malformatted requests will be accepted as given and no warnings will be generated about ignored or incorrect fields or values.

[data_parser/]v0.0.41[+fast][+prefer_refs]

Load data_parser/]v0.0.41 plugin to for formatting of data. Only compatible with openapi/slurmctld and openapi/slurmdbd content plugins.

+fast: Disable builtin warnings and other logic to strictly validate incoming requests. Should only ever be used in a production environment with very well tested clients and potentially malformatted requests will be accepted as given and no warnings will be generated about ignored or incorrect fields or values.

+prefer_refs: Prefer inline expansion of referenced schemas (via $ref) in generated OpenAPI specifications if the schema is only referenced once.

[data_parser/]v0.0.42[+fast][+minimize_refs][+inline_enums]

Load data_parser/]v0.0.42 plugin to for formatting of data. Only compatible with openapi/slurmctld and openapi/slurmdbd content plugins.

+fast: Disable builtin warnings and other logic to strictly validate incoming requests. Should only ever be used in a production environment with very well tested clients and potentially malformatted requests will be accepted as given and no warnings will be generated about ignored or incorrect fields or values.

+minimize_refs: Avoid inline expansion of referenced schemas (via $ref) in generated OpenAPI specifications if the schema is only referenced once.

+inline_enums: Avoid inline expansion of referenced schemas (via $ref) in generated OpenAPI specifications for enum arrays.

-f <file>

Read Slurm configuration from the specified file. See NOTES below.

--generate-openapi-spec

Dump JSON formatted OpenAPI specification to stdout and exit immediately. Loads miminal plugins required. Loading of slurm.conf(5) can be disabled by passing additional arguments -f /dev/null or setting SLURM_CONF=/dev/null in the environment.

-g <group id>

Change group id (and drop supplemental groups) before processing client request. This should be a unique group with no write access or special permissions. Do not set this to the group belonging to to SlurmUser or root or the daemon won't start with the default settings.

-h, --help

Help; print a brief summary of command options.

--max-connections <count>

Set the maximum number of connections to process at any one time. This is independent of the number of connections that can connect to slurmrestd at any one time. The kernel allows any number of connections to be pending for processing at any one time when SYN cookies are active.

Caution:: Each connection could cause one RPC to the controller daemons, leading to potential overloading of the controller. Each connection can also hold memory for the duration of the life of the connection. Having too many connections processing at once could use considerably more memory. Process limits (ulimit(3)) may require adjustment when this value is increased.
Default: 124

-s <plugin>[,plugin]...

Comma-delimited list of OpenAPI plugins to load, which will determine the available content. By default all available OpenAPI plugins will be loaded. See also -d for the data_parser plugins controlling the output format.

list: Display a list of the possible plugins to load.

[openapi/]slurmctld: Provides 'slurm/' endpoints for the loaded data_parser plugins.

[openapi/]slurmdbd: Provides 'slurmdb/' endpoints for the loaded data_parser plugins. This plugin will fail if accounting_storage/slurmdbd is not used in the cluster.

-t <THREAD COUNT>

Specify number of threads to use to process client connections. Ignored in inetd mode. Default: 20

-u <user id>

Change user id before processing client request. This should be a unique group with no write access or special permissions. Do not set this user to SlurmUser or root or the daemon won't start with the default settings.

-v

Verbose operation. Multiple v's can be specified, with each 'v' beyond the first increasing verbosity, up to 6 times (i.e. -vvvvvv). Higher verbosity levels will have significant performance impact.

-V

Print version information and exit.

ENVIRONMENT VARIABLES

The following environment variables can be used to override settings compiled into slurmrestd.

ABORT_ON_FATAL

When a fatal error is detected, use abort() instead of exit() to terminate the process. This allows backtraces to be captured without recompiling Slurm.

SLURM_CONF

The location of the Slurm configuration file.

SLURM_DEBUG_FLAGS

Specify debug flags for slurmrestd to use. See DebugFlags in the slurm.conf(5) man page for a full list of flags. The environment variable takes precedence over the setting in the slurm.conf.

SLURMRESTD_JSON

Control JSON serialization:

compact: Output JSON as compact as possible.

pretty: Output JSON in pretty format to make it more readable.

SLURM_JWT

This variable must be set to use JWT token authentication.

SLURMRESTD_AUTH_TYPES

Set allowed authentication types. See -a

SLURMRESTD_DEBUG

Set debug level explicitly. Valid values are 0-9, or the same string values as the debug options such as SlurmctldDebug in slurm.conf(5). Ignored if -v passed as argument during invocation.

SLURMRESTD_DATA_PARSER_PLUGINS

Comma-delimited list of data_parser plugins to load. See -d

SLURMRESTD_LISTEN

Comma-delimited list of host:port pairs or unix sockets to listen on.

SLURMRESTD_MAX_CONNECTIONS

Set the maximum number of connections to process at any one time. See --max-connections

SLURMRESTD_OPENAPI_PLUGINS

Comma-delimited list of OpenAPI plugins to load. See -s

SLURMRESTD_RESPONSE_STATUS_CODES

Comma-delimited list of OpenAPI method responses to generate in OpenAPI specification. Default: 200,default

SLURMRESTD_SECURITY

Control slurmrestd security functionality using the following comma-delimited values:

become_user: Allows slurmrestd to be run as root in order to become the requesting user for all requests. When combined with rest_auth/local, when a user connects via a named UNIX socket, slurmrestd will setuid()/setgid() into that user/group and then complete all requests as the given user. This mode is only intended for inet mode as the user change is permanent for the life of the process. This mode is incompatible with rest_auth/jwt and it is suggested to start slurmrestd with "-a rest_auth/local" arguments.

disable_unshare_files: Disables unsharing file descriptors with parent process.

disable_unshare_sysv: Disables unsharing the SYSV namespace.

disable_user_check: Disables check that slurmrestd is not running as root or SlurmUser, or with the root or SlurmUser's primary group.

SLURMRESTD_YAML

Control YAML serialization:

compact: Output YAML as compact as possible.

pretty: Output YAML in pretty format to make it more readable.

SIGNALS

SIGINT: slurmrestd will shutdown cleanly.

SIGPIPE: This signal is explicitly ignored.

NOTES

SPANK and clifilter plugins are not supported in slurmrestd due to their lack of thread safety. Active SPANK plugins and JobSubmitPlugins in slurmctld are independent of slurmrestd and can be used to enforce site policy on job submissions.

EXAMPLES

Generate OpenAPI schema without configuration

$ slurmrestd -f /dev/null  --generate-openapi-spec -s slurmdbd,slurmctld -d v0.0.42 > openapi.json

Start slurmrestd with a UNIX socket in listen mode:

$ export SLURMRESTD=/var/spool/slurm/restd/rest
$ slurmrestd -s slurmctld,slurmdbd -d v0.0.42 unix:$SLURMRESTD

Verify connectivity with slurmctld with a ping, with slurmrestd running in listen mode:

$ export $(scontrol token)
$ curl --unix-socket "${SLURMRESTD}" -H "X-SLURM-USER-TOKEN:${SLURM_JWT}" 'http://ignored_with_unix_sockets/slurm/v0.0.42/ping' | jq '.pings'
[
  {
    "hostname": "omicronpersei8",
    "pinged": "UP",
    "latency": 314,
    "mode": "primary"
  }
]

Verify connectivity with slurmdbd with a diag request, with slurmrestd running in listen mode:

$ export $(scontrol token)
$ curl --unix-socket "${SLURMRESTD}" -H "X-SLURM-USER-TOKEN:${SLURM_JWT}" 'http://ignored_with_unix_sockets/slurmdb/v0.0.42/diag' | jq '.pings'
1722009793

Query the status of a node with slurmrestd running in INETD mode:

$ echo -e "GET http://ignored/slurm/v0.0.42/node/host1 HTTP/1.1\r\n" | slurmrestd
HTTP/1.1 200 OK
Content-Length: 3174
Content-Type: application/json

{
  "nodes": [
    {
      "architecture": "x86_64",
      "burstbuffer_network_address": "",
      "boards": 1,
      "boot_time": {
        "set": true,
        "infinite": false,
        "number": 1720820315
      },
      "cluster_name": "",
      "cores": 16,
      "specialized_cores": 0,
      "cpu_binding": 0,
      "cpu_load": 446,
      "free_mem": {
        "set": true,
        "infinite": false,
        "number": 39871
      },
      "cpus": 32,
      "effective_cpus": 32,
      "specialized_cpus": "",
      "energy": {
        "average_watts": 0,
        "base_consumed_energy": 0,
        "consumed_energy": 0,
        "current_watts": {
          "set": false,
          "infinite": false,
          "number": 0
        },
        "previous_consumed_energy": 0,
        "last_collected": 0
      },
      "external_sensors": {},
      "extra": "",
      "power": {},
      "features": [],
      "active_features": [],
      "gpu_spec": "",
      "gres": "gpu:fake1:1(S:0),gpu:fake2:1(S:0)",
      "gres_drained": "N/A",
      "gres_used": "gpu:fake1:0(IDX:N/A),gpu:fake2:0(IDX:N/A)",
      "instance_id": "",
      "instance_type": "",
      "last_busy": {
        "set": true,
        "infinite": false,
        "number": 1722009794
      },
      "mcs_label": "",
      "specialized_memory": 0,
      "name": "host1",
      "next_state_after_reboot": [
        "INVALID"
      ],
      "address": "localhost",
      "hostname": "omicronpersei8",
      "state": [
        "IDLE"
      ],
      "operating_system": "Linux 6.5.0-44-generic #44-Ubuntu SMP PREEMPT_DYNAMIC Fri Jun  7 15:10:09 UTC 2024",
      "owner": "",
      "partitions": [
        "debug"
      ],
      "port": 5015,
      "real_memory": 127927,
      "res_cores_per_gpu": 0,
      "comment": "",
      "reason": "",
      "reason_changed_at": {
        "set": true,
        "infinite": false,
        "number": 0
      },
      "reason_set_by_user": "",
      "resume_after": {
        "set": true,
        "infinite": false,
        "number": 0
      },
      "reservation": "",
      "alloc_memory": 0,
      "alloc_cpus": 0,
      "alloc_idle_cpus": 32,
      "tres_used": "",
      "tres_weighted": 0.0,
      "slurmd_start_time": {
        "set": true,
        "infinite": false,
        "number": 1722009794
      },
      "sockets": 1,
      "threads": 2,
      "temporary_disk": 0,
      "weight": 1,
      "tres": "cpu=32,mem=127927M,billing=32,gres/gpu=2",
      "version": "24.11.0-0rc1"
    }
  ],
  "last_update": {
    "set": true,
    "infinite": false,
    "number": 1722010273
  },
  "meta": {
<<< TRIMMED >>>
  },
  "errors": [],
  "warnings": []
}

Submit a job to slurmrestd with it running in listen mode:

$ jq . example_job.json
{
  "job": {
    "script": "#!/bin/bash\nsleep 30",
    "name": "ExampleJob",
    "account": "sub1",
    "environment": [
      "PATH=/usr/bin/:/bin/"
    ],
    "current_working_directory": "/tmp/",
    "tasks": 12,
    "memory_per_cpu": 100,
    "time_limit": 240
  }
}

$ curl -H "Content-Type: application/json" --data-binary @example_job.json --unix-socket "${SLURMRESTD}" 'http://ignored/slurm/v0.0.42/job/submit'
{
  "job_id": 9,
  "step_id": "batch",
  "job_submit_user_msg": "",
  "meta": {
<<< TRIMMED >>>
  },
  "errors": [],
  "warnings": []
}

$ curl -H "Content-Type: application/json" --data-binary @example_job.json --unix-socket "${SLURMRESTD}" 'http://ignored/slurm/v0.0.42/job/submit'
{
  "job_id": 7,
  "step_id": "batch",
  "job_submit_user_msg": "",
  "meta": {
  },
  "errors": [],
  "warnings": [
    {
      "description": "Expected OpenAPI type=array (Slurm type=list) but got OpenAPI type=object (Slurm type=dictionary): {      "source": "#/job/environment/"
    }
  ]
}

COPYING

This file is part of Slurm, a resource management program. For details, see <https://slurm.schedmd.com/>.

Slurm is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

Slurm is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

This document was created by man2html using the manual pages.
Time: 19:39:59 GMT, May 07, 2025