1slurmd(8) Slurm Daemon slurmd(8)
2
6 slurmd - The compute node daemon for Slurm.
7
10 slurmd [OPTIONS...]
11
14 slurmd is the compute node daemon of Slurm. It monitors all tasks run‐
15 ning on the compute node , accepts work (tasks), launches tasks, and
16 kills running tasks upon request.
17
20 -b Report node rebooted when daemon restarted. Used for testing
21 purposes.
22 -c Clear system locks as needed. This may be required if slurmd
24 terminated abnormally.
25 -C Print the actual hardware configuration (not the configuration
27 from the slurm.conf file) and exit. The format of output is the
28 same as used in slurm.conf to describe a node's configuration
29 plus its uptime.
30 --conf <node parameters>
32 Used in conjunction with the -Z option. Used to override or de‐
33 fine additional parameters of a dynamic node using the same syn‐
34 tax and parameters used to define nodes in the slurm.conf. Spec‐
35 ifying any of CPUs, Boards, SocketsPerBoard, CoresPerSocket or
36 ThreadsPerCore will override the defaults defined by the -C op‐
37 tion. NodeName and Port are not supported.
38 For example if slurmd -C reports
40 NodeName=node1 CPUs=16 Boards=1 SocketsPerBoard=1 CoresPerSocket=8 ThreadsPerCore=2 RealMemory=31848
41 the following --conf specifications will generate the corre‐
43 sponding node definitions:
44 --conf "Gres=gpu:2"
45 NodeName=node1 CPUs=16 Boards=1 SocketsPerBoard=1 CoresPerSocket=8 ThreadsPerCore=2 RealMemory=31848 Gres=gpu:2
46 --conf "RealMemory=30000"
48 NodeName=node1 CPUs=16 Boards=1 SocketsPerBoard=1 CoresPerSocket=8 ThreadsPerCore=2 RealMemory=30000
49 --conf "CPUs=16"
51 NodeName=node1 CPUs=16 RealMemory=331848
52 --conf "CPUs=16 RealMemory=30000 Gres=gpu:2"
54 NodeName=node1 CPUs=16 RealMemory=30000 Gres=gpu:2"
55 --conf-server <host>[:<port>]
57 Comma-separated list of controllers, the first being the primary
58 slurmctld. A port can (optionally) be specified for each con‐
59 troller. These hosts are where the slurmd will fetch the config‐
60 uration from when running in "configless" mode.
61 -d <file>
63 Specify the fully qualified pathname to the slurmstepd program
64 to be used for shepherding user job steps. This can be useful
65 for testing purposes.
66 -D Run slurmd in the foreground. Error and debug messages will be
68 copied to stderr.
69 -f <file>
71 Read configuration from the specified file. See NOTES below.
72 -F[feature]
74 Start this node as a Dynamic Future node. It will try to match a
75 node definition with a state of FUTURE, optionally using the
76 specified feature to match the node definition.
77 -G Print Generic RESource (GRES) configuration (based upon
79 slurm.conf GRES merged with gres.conf contents for this node)
80 and exit.
81 -h Help; print a brief summary of command options.
83 -L <file>
85 Write log messages to the specified file.
86 -M Lock slurmd pages into system memory using mlockall (2) to dis‐
88 able paging of the slurmd process. This may help in cases where
89 nodes are marked DOWN during periods of heavy swap activity. If
90 the mlockall (2) system call is not available, an error will be
91 printed to the log and slurmd will continue as normal.
92 It is suggested to set LaunchParameters=slurmstepd_memlock in
94 slurm.conf(5) when setting -M.
95 -n <value>
97 Set the daemon's nice value to the specified value, typically a
98 negative number. Also note the PropagatePrioProcess configura‐
99 tion parameter.
100 -N <nodename>
102 Run the daemon with the given nodename. Used to emulate a larger
103 system with more than one slurmd daemon per node. Requires that
104 Slurm be built using the --enable-multiple-slurmd configure op‐
105 tion.
106 -s Change working directory of slurmd to SlurmdLogFile path if pos‐
108 sible, or to SlurmdSpoolDir otherwise. If both of them fail it
109 will fallback to /var/tmp.
110 -v Verbose operation. Multiple -v's increase verbosity.
112 -V, --version
114 Print version information and exit.
115 -Z Start this node as a Dynamic Normal node. If no --conf is speci‐
117 fied, then the slurmd will register with the same hardware con‐
118 figuration as defined by the -C option.
119
121 The following environment variables can be used to override settings
122 compiled into slurmd.
123 SLURM_CONF The location of the Slurm configuration file. This
126 is overridden by explicitly naming a configuration
127 file on the command line.
128 SLURM_DEBUG_FLAGS Specify debug flags for slurmd to use. See De‐
130 bugFlags in the slurm.conf(5) man page for a full
131 list of flags. The environment variable takes
132 precedence over the setting in the slurm.conf.
133
135 SIGTERM SIGINT
136 slurmd will shutdown cleanly, waiting for in-progress rollups to
137 finish.
138 SIGHUP Reloads the slurm configuration files, similar to 'scontrol re‐
140 configure'.
141 SIGUSR2
143 Reread the log level from the configs, and then reopen the log
144 file. This should be used when setting up logrotate(8).
145 SIGPIPE
147 This signal is explicitly ignored.
148
150 If slurmd is started with the -D option then the core file will be
151 written to the current working directory. Otherwise if SlurmdLogFile
152 is a fully qualified path name (starting with a slash), the core file
153 will be written to the same directory as the log file. Otherwise the
154 core file will be written to the SlurmSpoolDir directory, or
155 "/var/tmp/" as a last resort. If none of the above directories can be
156 written, no core file will be produced.
157
160 It may be useful to experiment with different slurmd specific configu‐
161 ration parameters using a distinct configuration file (e.g. timeouts).
162 However, this special configuration file will not be used by the slurm‐
163 ctld daemon or the Slurm programs, unless you specifically tell each of
164 them to use it. If you desire changing communication ports, the loca‐
165 tion of the temporary file system, or other parameters used by other
166 Slurm components, change the common configuration file, slurm.conf.
167 If you are using configless mode with a login node that runs a lot of
169 client commands, you may consider running slurmd on that machine so it
170 can manage a cached version of the configuration files. Otherwise, each
171 client command will use the DNS record to contact the controller and
172 get the configuration information, which could place additional load on
173 the controller.
174
177 Copyright (C) 2002-2007 The Regents of the University of California.
178 Copyright (C) 2008-2010 Lawrence Livermore National Security. Copy‐
179 right (C) 2010-2022 SchedMD LLC. Produced at Lawrence Livermore Na‐
180 tional Laboratory (cf, DISCLAIMER).
181 This file is part of Slurm, a resource management program. For de‐
183 tails, see <https://slurm.schedmd.com/>.
184 Slurm is free software; you can redistribute it and/or modify it under
186 the terms of the GNU General Public License as published by the Free
187 Software Foundation; either version 2 of the License, or (at your op‐
188 tion) any later version.
189 Slurm is distributed in the hope that it will be useful, but WITHOUT
191 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
192 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
193 for more details.
194
197 /etc/slurm.conf
198
201 slurm.conf(5), slurmctld(8)
202August 2022 Slurm Daemon slurmd(8)