1burst_buffer.conf(5) Slurm configuration file burst_buffer.conf(5)
2
6 burst_buffer.conf - Slurm configuration file for burst buffer manage‐
7 ment.
8
11 burst_buffer.conf is an ASCII file which describes the configuration of
12 burst buffer resource management. This file is only required on the
13 head node(s), where the slurmctld daemon executes. The file location
14 can be modified at system build time using the DEFAULT_SLURM_CONF
15 parameter or at execution time by setting the SLURM_CONF environment
16 variable. The file will always be located in the same directory as the
17 slurm.conf file. In order to support multiple configuration files for
18 multiple burst buffer plugins, the configuration file may alternately
19 be given a name containing the plugin name. For example, if
20 "burst_buffer.conf" is not found, the burst_buffer/generic configura‐
21 tion could be read from a file named "burst_buffer_generic.conf".
22 Parameter names are case insensitive. Any text following a "#" in the
24 configuration file is treated as a comment through the end of that
25 line. Changes to the configuration file take effect upon restart of
26 Slurm daemons, daemon receipt of the SIGHUP signal, or execution of the
27 command "scontrol reconfigure".
28 The configuration parameters available include:
30 AllowUsers
33 Comma separated list of user names and/or IDs permitted to use
34 burst buffers. The options AllowUsers and DenyUsers can not
35 both be specified. By default all users are permitted to use
36 burst buffers. Possible future future enhancement for generic
37 plugin.
38 AltPool
41 The AltPool value consists of a comma separated list of alter‐
42 nate pools of resources, each of which includes a name separated
43 by a colon and a numeric value. The numeric value can include a
44 suffix of "k", "m", "g", "t" and "p" which multiplies the
45 numeric value by 1,024, 1,048,576, 1,073,741,824 etc. The
46 numeric value is a 64-bit value. for example, "Alt‐
47 Pool=test:20g,small:4g
48 CreateBuffer
51 Fully qualified path name of a program which will create both
52 persistent and per-job burst buffers. This option is not used
53 by the burst_buffer/cray plugin.
54 DefaultPool
57 Name of the pool used by default for resource allocations. The
58 default value is the first pool reported by the burst buffer in‐
59 frastructure. This option is only used by the burst_buffer/cray
60 plugin.
61 DenyUsers
64 Colon delimited list of user names and/or IDs prevented from
65 using burst buffers. The options AllowUsers and DenyUsers can
66 not both be specified. By default all users are permitted to
67 use burst buffers.
68 DestroyBuffer
71 Fully qualified path name of a program which will destroy both
72 persistent and per-job burst buffers. This option is not used
73 by the burst_buffer/cray plugin.
74 Flags String used to control various functions. Multiple options may
77 be comma separated. Supported options include:
78 DisablePersistent
80 Prevents regular users from being able to create and
81 destroy persistent burst buffers. This is the default
82 behaviour, only privileged users (Slurm operators and
83 administrators) can create or destroy persistent burst
84 buffers.
85 EmulateCray
87 Emulating a Cray DataWarp system using the dw_wlm_cli
88 script in the burst_buffer/cray plugin.
89 EnablePersistent
91 Enables regular users to create and destroy persistent
92 burst buffers. By default, only privileged users (Slurm
93 operators and administrators) can create or destroy per‐
94 sistent burst buffers.
95 PrivateData
97 If set, then only Slurm operators and the burst buffer
98 owner can see burst buffer data.
99 SetExecHost
101 If using Cray DataWarp version CLE.6.0.UP07 or later,
102 this flag may be set to notify DataWarp of the login node
103 for an interactive jobs. This changes behavior such that
104 the login node will be given access to the DataWarp burst
105 buffers.
106 TeardownFailure
108 If set, then teardown a burst buffer after file staging
109 error. Otherwise preserve the burst buffer for analysis
110 and manual teardown.
111 GetSysState
114 Fully qualified path name of a program which will return the
115 current burst buffer state. See the src/plugins/burst_buf‐
116 fer/generic/bb_get_state.example in the Slurm distribution for
117 an example. For the Cray plugin, this should be the path of the
118 dw_wlm_cli command and it's default value is
119 /opt/cray/dw_wlm/default/bin/dw_wlm_cli.
120 GetSysStatus
123 Fully qualified path name of a program which will return the
124 current burst buffer status. For the Cray plugin, this should
125 be the path of the dwstat command and it's default value is
126 /opt/cray/dws/default/bin/dwstat.
127 Granularity
130 Granularity of job space allocations in units of bytes. The
131 numeric value may have a suffix of "m" (megabytes), "g" (giga‐
132 bytes), "t" (terabytes), "p" (petabytes), or "n" (nodes). Bytes
133 is assumed if no suffix is supplied. This option is not used by
134 the burst_buffer/cray plugin.
135 OtherTimeout
138 If a burst buffer operation (other than job validation, stage
139 in, or stage out) runs for longer than this number of seconds,
140 the job will be placed in a held state. A Slurm administrator
141 will be required to release the job. By default there is a 300
142 second (5 minute) timeout for these operations. Also see
143 StageInTimeout, StageOutTimeout, and ValidateTimeout options.
144 PrivateData
147 If set to "true" then users will only be able to view burst buf‐
148 fers they can use. Slurm administrators will still be able to
149 view all burst buffers. By default, users can view all burst
150 buffers.
151 StageInTimeout
154 If the stage in of files for a job takes more than this number
155 of seconds, the burst buffer will be released and the job will
156 be placed in a held state. A Slurm administrator will be
157 required to release the job. By default there is a one day
158 timeout for the stage in process.
159 StageOutTimeout
162 If the stage out of files for a job takes more than this number
163 of seconds, the burst buffer will be released and the job will
164 be purged. By default there is a one day timeout for the stage
165 out process.
166 StartStageIn
169 Fully qualified path name of a program which will stage files in
170 for a job. See the src/plugins/burst_buf‐
171 fer/generic/bb_start_stage_in.example in the Slurm distribution
172 for an example. This option is not used by the burst_buf‐
173 fer/cray plugin.
174 StartStageOut
177 Fully qualified path name of a program which will stage files
178 out for a job. See the src/plugins/burst_buf‐
179 fer/generic/bb_start_stage_out.example in the Slurm distribution
180 for an example. This option is not used by the burst_buf‐
181 fer/cray plugin.
182 StopStageIn
185 Fully qualified path name of a program which will stop staging
186 files in for a job. See the src/plugins/burst_buf‐
187 fer/generic/bb_stop_stage_out.example in the Slurm distribution
188 for an example. This option is not used by the burst_buf‐
189 fer/cray plugin.
190 StopStageOut
193 Fully qualified path name of a program which will stop staging
194 files in for a job. See the src/plugins/burst_buf‐
195 fer/generic/bb_stop_stage_out.example in the Slurm distribution
196 for an example. This option is not used by the burst_buf‐
197 fer/cray plugin.
198 ValidateTimeout
201 If the validation of a job submission request takes more than
202 this number of seconds, the submission will be rejected. The
203 value of ValidateTimeout must be less than the value of Message‐
204 Timeout configured in the slurm.conf file or job submission
205 requests may fail with a response timeout error. By default
206 there is a 5 second timeout for the validation operations.
207
210 ##################################################################
211 # Slurm's burst buffer configuration file (burst_buffer.conf)
212 ##################################################################
213 AllowUsers=alan,brenda
214 PrivateData=true
215 Gres=nodes:10,other:20
216 #
217 Granularity=1G
218 #
219 StageInTimeout=30 # Seconds
220 StageOutTimeout=30 # Seconds
221 #
222 CreateBuffer=/usr/local/slurm/17.11/sbin/CB
223 DestroyBuffer=/usr/local/slurm/17.11/sbin/DB
224 GetSysState=/usr/local/slurm/17.11/sbin/GSS
225 StartStageIn=/usr/local/slurm/17.11/sbin/SSI
226 StartStageOut=/usr/local/slurm/17.11/sbin/SSO
227 StopStageIn=/usr/local/slurm/17.11/sbin/PSI
228 StopStageOut=/usr/local/slurm/17.11/sbin/PSO
229
232 Copyright (C) 2014-2018 SchedMD LLC.
233 This file is part of Slurm, a resource management program. For
235 details, see <https://slurm.schedmd.com/>.
236 Slurm is free software; you can redistribute it and/or modify it under
238 the terms of the GNU General Public License as published by the Free
239 Software Foundation; either version 2 of the License, or (at your
240 option) any later version.
241 Slurm is distributed in the hope that it will be useful, but WITHOUT
243 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
244 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
245 for more details.
246
249 slurm.conf(5)
250burst_buffer.conf 18.08 June 2018 burst_buffer.conf(5)