/[pdpsoft]/nl.nikhef.pdp.dynsched-pbs-plugin/trunk/RELEASE
ViewVC logotype

Diff of /nl.nikhef.pdp.dynsched-pbs-plugin/trunk/RELEASE

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 2026 by templon, Fri Oct 8 13:11:04 2010 UTC revision 2027 by templon, Fri Oct 8 14:49:56 2010 UTC
# Line 1  Line 1 
1  This file contains release notes and a change history for the  This file contains release notes and a change history for the
2  lcg-info-dynamic-scheduler information provider.  pbs/torque and maui plugins for the lcg-info-dynamic-scheduler information provider.
3  It also contains release notes and a change history for the  The notes are most recent first.
 PBS/Torque/Maui backend commands.  The notes are  
 most recent first.  
   
 Release 2.2.1  
    The dynamic scheduler was changed to cease printing the GlueCEAccessControlBaseRule.  
    2.2.0 did not work since GIP considers all changes to multivalued attributes (like  
   ACBRs) to be significant.  
4    
5  Release 2.2.0  Release 2.2.0
6     The dynamic scheduler was changed in order to deal with the DENY     Moved to new repository format and changed RPM name.  Rather large changes in two areas.  First, in the logFileParser
7     tags being used in the short-term solution (June 2007) for job priorities.     part of the pbsServer.py code.  The new "account" field printed in the torque accounting logs broke the old parser,
8     The dynamic scheduler does the following with ACBRs placed on VOViews:     this is fixed in the current release.  The second set of changes have to do with how the LiveServer code parses
9        - it discards any ACBR that does not begin with either "VO:" or "VOMS:"     the 'qstat -f' output.  When present (recent versions of torque), the code now uses 'startTime' as the start time
10        - if there is more than one ACBR left in the list, it only uses the last one     of the job, and uses the 'euser' and 'egroup' as the user and group of the job.  If these fields are not present,
11           in the list, and prints a warning message to standard error and to syslog     the code reverts to the old behavior.
12        - it allows multiple DENY tags  
13        - there is no checking on the consistency between the ACBR and DENY tags in a view.     Two user-visible changes: there are two new job attributes in the "Job" object.  One is the walltime used by the job,
14       the other is the 'startAnchor' field which tells you whether the code has used the 'startTime' to find the start time,
15       or has done it the old way by subtracting the walltime used from the current time.
16    
17    Information below this point refers to older versions of the package, under the older naming
18    scheme.
19    
20  Release 2.1.0  Release 2.1.0
21            - Remove support for specifying hostname on command line (savannah bug 35662)
22    
23     lrms.py was changed in order to support caching of search results.  Release 2.0.1
24     Most of the time spent in lcg-info-dynamic-scheduler was due to          - fix LICENSE, otherwise no change.
    queries like "find all jobs from group 'lhcb', in state 'waiting',  
    for queue 'qlong'.  Queries like this are now cached for future use,  
    and can also be supplied *before* use, like they now are for  
    lcg-info-dynamic-scheduler.  That program now generates slices  
    of the job list for the various combinations queue/group/state  
    that will be needed while running the program.  
   
    There were previously two different 'return a list of matching job'  
    functions, with different interfaces.  These now have a unified  
    interface so that result caching can be supported.  This does break  
    backwards compatibility for lrms.py.  
25    
26  Release 2.0.0  Release 2.0.0
27            - uses new lrms.py, otherwise no changes.
    Rather massive changes in parsing logic, to be able to handle VOViews  
    with VOMS FQANs.  
   
    VOMS FQANs are handled both by the input routines, which know what  
    to do with them when reading the static LDIF file, as well as  
    the group mapping logic, that knows how to associate FQANs  
    with unix groups.  To this end, the vomap construct in the  
    lcg-info-dynamic-scheduler config file now supports lines like  
   
            lhcbsgm:/VO=lhcb/GROUP=/lhcb/ROLE=lcgadmin  
   
    in addition to the original lines like  
   
            atlgrid : atlas  
   
    which would map group 'atlgrid' to "VO : atlas".  
   
    The parsing of the GlueCEUniqueID and GlueVOView blocks has  
    also changed rather drastically, so that previous problems with  
    numbers, dashes, etc in queue names and hostnames are no longer a  
    problem.  Instead of parsing the GlueCEUniqueID field to get the  
    queue name, the program now reads GlueCEName and uses that for  
    the queue name.  
   
    Also, the file vomaxjobs-generic (documentation) was added,  
    and the rest of the documentation and example files was  
    substantially upated for the new release.  
   
    Otherwise no changes since 1.6.3.  
   
    For people using the test suite: the versions of the test output  
    included in 2.0.0 will cause tests of older versions to fail.  This  
    is unavoidable since the old parsing logic was based on the order  
    in which blocks appeared in the ldif file, while the new version  
    uses python 'dicts' which have an unpredictable order when  
    iterated.  To make the order predictable (for purposes of test  
    harness), the keys are sorted before the program starts to print.  
    The older versions do not sort the output before printing, hence  
    tests of the old versions with the new files will fail.  
   
 Release 1.6.3  
   
         Fix for GGUS bug 10155 -- had to do with YAIM adding unnecessary lines like  
             alice:alice  
         to the [vomap] stanza.  The program did not expect to get these lines so it  
         of course did something rather silly with them, resulting in the behavior  
         reported in the GGUS bug.  
28    
29  Release 1.6.1  Release 1.6.1
30            - update documentation and examples to be consistent with 2.0 release
         Bug fix for lcg-info-dynamic-scheduler; fix regexp matching  
         GlueCEUniqueID.  the regexp in 1.6.0 missed  
            - CEs with a "-" character in the hostname  
            - queue names with underscores, uppercase letters, and numbers  
   
         There are examples of each of these classes on the production system,  
         so this upgrade is critical.  
31    
32  Release 1.6.0  Release 1.6.0
33    
         - changes to parsing of static LDIF file to pick up gLite CEs with "blah"  
            instead of "jobmanager".  Note this is largely untested!!  
         - added test suite to prevent bug regression  
34          - some changes to build system (three targets increases aggravation)          - some changes to build system (three targets increases aggravation)
35          - some changes to pbsServer classes to assist in debugging.          - some changes to pbsServer classes to assist in debugging.
36          - some changes to vomaxjobs-maui to assist in debugging/testing;          - some changes to vomaxjobs-maui to assist in debugging/testing;
# Line 122  Release 1.5.2: Line 53  Release 1.5.2:
53          pbs package: Fix to vomaxjobs-maui to deal with cases where there is          pbs package: Fix to vomaxjobs-maui to deal with cases where there is
54          extra 'warning' output near the top of the command output from diagnose -g.          extra 'warning' output near the top of the command output from diagnose -g.
55    
         generic package: fix bug with logging; undefined variable caused fatal program  
         exit while trying to print warning message.  
   
56  Release 1.5.1:  Release 1.5.1:
57    
58          fix dependency problems with RPMs.          fix dependency problems with RPMs.
59    
60  Release 1.5.0  Release 1.5.0
61    
62  * add RELEASE (this file) to docs dir in generic package RPM          in vomaxjobs-maui, adapt to handle MAXPROC specifications like
63            MAXPROC=soft,hard The code reports the 'hard' limit, since
64  * Minor change to build system to make tag events in ChangeLogs          this is relevant when the system is not full, and this is when
65     easier to read.          it's needed.  Maui uses the soft limit on a full system, but
66            in this case the info provider will drop FreeSlots to zero as
67  * lcg-info-dynamic-scheduler:          soon as jobs remain in the queued state instead of executing
68            immediately.
   - It is possible (e.g. by dramatically reducing MAXPROC config in Maui) for  
      a VO to have more running jobs in the LRMS than allowed by MAXPROC.  
      In this case a negative value was reported for FreeSlots.  Fixed.  
   
   - implemented logging to syslog  
   
 * vomaxjobs-maui:  
   
   - adapt to handle MAXPROC specifications like MAXPROC=soft,hard  
      The code reports the 'hard' limit, since this is relevant when the  
      system is not full, and this is when it's needed.  Maui uses the  
      soft limit on a full system, but in this case the info provider will  
      drop FreeSlots to zero as soon as jobs remain in the queued state  
      instead of executing immediately.  
   
 Release 1.4.3  
   
 * lcg-info-dynamic-scheduler:  
   
   - fix for Savannah bug 14946: overflow of conversion of response time  
      values from float (internal) to int (output representation).  Now prints the  
      magic value of 2146060842 as an upper limit.  
69    
70  Release 1.4.2  Release 1.4.2
71    
72  * pbsServer.py:          in pbsServer.py: included Steve Traylen's patch to deal with jobs for which the
73            uid/gid printed by 'qstat' is not listed in the in running machine's
74    - included Steve Traylen's patch to deal with jobs for which the          pw DB.  This can happen when the CE is not the same physical
75       uid/gid printed by 'qstat' is not listed in the in running machine's          machine as the actual LRMS server.
      pw DB.  This can happen when the CE is not the same physical  
      machine as the actual LRMS server.  
   
76    
77  Estimated Response Time Info Providers (v 1.4.1)  Estimated Response Time Info Providers (v 1.4.1)
78  ------------------------------------------------  ------------------------------------------------
# Line 203  the requirements on the backend commands Line 106  the requirements on the backend commands
106  provide in the case that you are using a different  provide in the case that you are using a different
107  scheduler or LRMS.  Tim Bell at CERN can help for people  scheduler or LRMS.  Tim Bell at CERN can help for people
108  using LSF.  using LSF.
   
   
   

Legend:
Removed from v.2026  
changed lines
  Added in v.2027

grid.support@nikhef.nl
ViewVC Help
Powered by ViewVC 1.1.28