SLURM.pm untested

filter: Statementstmt 0.0% 0 / 1 Branchbran n/a Conditioncond n/a MC/DCmcdc n/a Subroutinesub n/a Podpod n/a totaltotal 0.0% 0 / 1 SCAR -
1 package HPCD::SLURM;
2 ### HPCD::SLURM.pm #########################################################################
3  
4 =head1 NAME
5  
6 HPCD::SLURM
7  
8 =head1 VERSION
9  
10 Version 0.01
11  
12 =cut
13  
14 our $VERSION = '0.01';
15  
16 =head1 SYNOPSIS
17  
18 use HPCI;
19  
20 my $group = HPCI->group(cluster => 'SLURM', ...);
21  
22 =head1 DESCRIPTION
23  
24 This is a driver module that can be used internally from HPCI, by
25 specifying a cluster type of 'SLURM'. It customizes the HPCI access
26 interface to control jobs submitted to an SLURM (Simple Linux Utility
27 for Resource Management) cluster.
28  
29 The differences provided by the SLURM driver are:
30  
31 =over
32  
33 =item -
34  
35 the following resource names have slurm-specific aliases
36  
37 =over
38  
39 =item h_time
40 time
41  
42 =back
43  
44 Using these aliases causes your code to be SLURM-specific. That makes
45 converting it to use an alternate type of cluster more painful. However,
46 if you do not expect to ever be moving your code to a different cluster
47 type, this might simplify the interface for people who already know the
48 SLURM-specific resource names and wish to avoid the confusion that might
49 come from using the generic names.
50  
51 =item -
52  
53 a stage attribute B<native_args_string> is available which specifies a string of
54 argument(s) that will be included in the srun/sbatch command that starts execution
55 of that stage
56  
57 =item -
58  
59 The result information hash returned by C<$stats = $group-E<gt>execute>
60 includes all of the info provided by the SLURM sacct command, not just
61 the exit status.
62  
63 =item -
64  
65 If a stage is terminated because it exceeds the requested memory resource
66 limit, then it will normally be retried using a higher limit.
67 The default is to first try with '2G' (if no mem resource limit
68 is explicitly specified). After mem failures, the next larger size in the
69 default sequence (2G 4G 8G 16G 32G) is attempted.
70  
71 =back
72  
73 =head1 SEE ALSO
74  
75 =over
76  
77 =item HPCI
78  
79 Describes the generic HPCI iterface - this manual is only providing
80 the exceptions to that document (and its related documents).
81  
82 =item HPCI::Group
83  
84 Describes the interface common to all B<HPCI Group>
85 objects, regardless of the particular type of cluster that
86 is actually being used to run the stages.
87  
88 =item HPCI::Stage
89  
90 Describes the interface common to stage object returned
91 by all B<HPCI Stage> objects, regardless of the
92 particular type of cluster that is actually being used to
93 run the stages.
94  
95 =back
96  
97 =head1 AUTHOR
98  
99 Christopher Lalansingh - Boutros Lab
100  
101 John Macdonald - Boutros Lab
102  
103 Anqi (Joyce) Yang - Boutros Lab
104  
105 =head1 ACKNOWLEDGEMENTS
106  
107 Paul Boutros, Phd, PI - Boutros Lab
108  
109 The Ontario Institute for Cancer Research
110  
111 =cut
112  
113 1;
114