1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
|
SGI kernel level checkpointing
------------------------------
___INFO__MARK_BEGIN__
-------------------------------------------------------------------------
The Contents of this file are made available subject to the terms of
the Sun Industry Standards Source License Version 1.2
Sun Microsystems Inc., March, 2001
Sun Industry Standards Source License Version 1.2
=================================================
The contents of this file are subject to the Sun Industry Standards
Source License Version 1.2 (the "License"); You may not use this file
except in compliance with the License. You may obtain a copy of the
License at http://gridengine.sunsource.net/Gridengine_SISSL_license.html
Software provided under this License is provided on an "AS IS" basis,
WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING,
WITHOUT LIMITATION, WARRANTIES THAT THE SOFTWARE IS FREE OF DEFECTS,
MERCHANTABLE, FIT FOR A PARTICULAR PURPOSE, OR NON-INFRINGING.
See the License for the specific provisions governing your rights and
obligations concerning the Software.
The Initial Developer of the Original Code is: Sun Microsystems, Inc.
Copyright: 2001 by Sun Microsystems, Inc.
All Rights Reserved.
-------------------------------------------------------------------------
___INFO__MARK_END__
The SGI CPR kernel-level checkpointing interfaces use the SGI cpr(1)
command to checkpoint and restart Grid Engine/Grid Engine Enterprise
Edition jobs. Here is some specific information about the SGI CPR
checkpointing interface scripts.
1. The cpr(1) command does not support waiting for a restarted batch job to
complete. This means that the o2k_restart_command script must itself wait
until the job has completed. It does this by polling in a loop using ps(1)
and sleep(1) until the job script exits. Currently, there is no method
to get the status of the restarted job.
2. When setting up a checkpointing interface for these scripts in the
Qmon checkpoint config dialog, the checkpoint commands should be
identified in the following format. The names in <> should be filled
in with the appropriate values.
<sge_root>/ckpt/<chkpnt_script_name> $job_id $job_pid $ckpt_dir
3. The checkpoint interface scripts log some general information to the file
$ckpt_dir/ckpt.log
4. The checkpoint files are placed in $ckpt_dir/ckpt.$job_id. A detailed
log is available in the user's home directory under the name
$REQNAME.co$JOD_ID.
5. These scripts are designed to handle a system failure during checkpointing.
That is, the latest checkpoint file will not be used for a restart until the
checkpoint command completes. If the system fails before the checkpoint has
completed, the last checkpoint file will be used for restart.
6. The checkpoint files will not be deleted in the cprcray_clean_command script
if either the SGE_LEAVE_CKPT_DIR environment variable is set to a non-null
string or if the fourth parameter in the clean command for the checkpoint
configuration is set to the string "save".
7. The cpr_ckpt_command and cpr_migration_command contain a qalter command to
force the job to restart on the same host as it was checkpointed.
8. These scripts checkpoint based on the ASH if it is available, otherwise
they use the process group ID.
9. These scripts are designed to work with IRIX 6.5 or later or IRIX 6.4 with
patch 3453 installed.
|
