root/gip/tags/gip_1_1_0/changelog

Revision 2946, 4.8 kB (checked in by burt, 8 months ago)

--

Line 
1 Changelog
2
3 == 1.1.0 ==
4 * Eliminate erroneous publication of 'waiting' queue for SGE
5
6 == 1.1.0rc3 ==
7 * Fixes for SGE (at least we should be able to get some basic information correctly published now)
8 * Fixes for SE publication (including not publishing a SE at all and publishing the CE as a gsiftp server)
9 * Some enhancements to be in-line with the WLCG Installed Capacity document
10 * Better error handling (missing/unreadable config.ini, missing cluster)
11 * Better reporting of PBS queues and free slots
12 * Fixes for reporting site sponsorship
13 * Increased timeout from 60s to 240s for each provider
14 * New GIP version provider
15        
16 == 1.1.0rc2 ==
17 * Bugfix: Incorrect foreign keys for vdt-version and timestamp info (18)
18 * Bugfix: a classic SE published different GlueSEUniqueIDs in the GlueSEUniqueID stanza and the GlueCSEBind stanza
19 * Handle advertise_se parameter in config.ini
20 * Pretty_gip_info fixes
21 * New code for the MyOSG plugin (OSG GOC)
22        
23 == 1.1.0rc1 ==
24 * More fleshed out (but probably still broken) SGE provider
25 * Fix to include default SE and publish correctly
26 * Revert to config.ini instead of current-config.ini
27        
28 == 1.1.0 ==
29 * Removal of gip-attributes.conf and configure_gip perl script, because:
30 * We now read config.ini directly
31 * Reconfigured GIP providers to determine the correct batch system and storage
32   system on-the-fly, removing the need for admin configuration
33 * We do not attempt to auto-detect subclusters
34 * New wrapper script with new name (osg_info_wrapper.py): (CEMon
35   compatibility issue)
36 * Wrapper script supports a new delete-attributes.conf
37 * Completely new providers for Condor, LSF, and SGE.  SGE and LSF are lightly
38   tested (SGE probably does not yet work out of the box).
39 * Condor provider now advertises data consistently with the other batch
40   system providers.
41 * New BestMan provider
42 * New dCache provider
43 * New provider for dCache 1.9 which does NOT need the dCache admin interface
44 * Improved adherence to Glue 1.3
45 * Increased unit test coverage
46 * Glue parameters CESEbind, site, subclusters, and cluster entries are all
47   dynamic
48 * Improved and more secure method for advertising VO-installed software
49 * Glue 1.3 style AccessBaseControlRules are used (CEMon compatibility issue)
50 * Storage and subcluster configurations via config.ini have been refactored
51   to be simpler and more straightforward.
52
53        
54 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
55 == 1.0.8 ==
56 * Fix missing and extraneous imports in token_info_provider.py
57
58 == 1.0.7 ==
59 * In configure_gip, check for globus only when looking for LSF queues
60        
61 == 1.0.6 ==
62 * Include plugin to calculate GlueCEStateEstimatedResponseTime ("ERT")
63 * Include logging for ERT
64 * Allow attributes in the DN to have internal whitespace
65        
66 == 1.0.5 ==
67 * Fix syntax error (spurious backticks) in configure_gip
68        
69 == 1.0.4 ==
70 * Bugfix for GlueVOInfoLocalID, which needs to include GlueSEUniqueID.
71
72 == 1.0.3 ==
73 * Sets reasonable defaults for GlueCEPolicyMaxObtainableWallClockTime, GlueCEPolicyMaxObtainableCPUTime, GlueCEStateWorstResponseTime
74
75 == 1.0.2 ==
76
77 Bugs fixed:
78 * Respect max_queueable for PBS (10)
79 * Multiple SE per CESE bindgroup (11)
80 * Schema compliance for GlueServiceStatusInfo?, GlueSARetentionPolicy,
81   GlueSAAccessLatency, GlueSAExpirationMode,GlueSACapability,
82   GlueSEImplementationName,GlueSEStatus (9,12)
83 * Custom attributes via alter-attributes (14)
84 * PBS Default wallclock set to 6000 instead of 0
85
86 Features added:
87 * Logging (13)
88 * LSF queues parsed from RVF instead of bqueues (thanks Alan Sill)
89
90 == 1.0.1 ==
91
92 Bugs fixed (references to GIP ticket numbers in [])
93   - The dynamic providers were parsing the VO list differently than in
94     previous cases [1]
95   - For an SRM SE, GlueSEArchitecture is changed from "disk" to "other".
96     This avoids the classicSE script trying to execute df in the wrong
97     places [2, 8]
98   - $GIP_LOCATION/etc/gip.conf was not being read in some cases due to
99     invalid path expansion [3]
100   - Multiple collector hosts were not getting parsed correctly (GIP
101     ticket [4]
102   - OSG Version is now published along with the software tags [5]
103   - PBS dynamic provider was not parsing waiting jobs correctly (GIP
104     ticket [6]
105   - For non-PBS sites using dynamic dcache, GlueSEUniqueID was not being
106     correctly generated [7]
107   - GlueCESEBindCEAccessPoint will not double-count the GlueSAPath now
108   - Wrong permissions and ownership were generated for
109     dcache_storage.conf, gip.conf
110   - CESEBind now advertises /UNAVAILABLE instead of dropping the
111     attribute if the access point is not set
112   - SRM hostname was being advertised based on reverse DNS, which is a
113     problem for sites where the SRM host is a CNAME instead of an A
114     name.  We now print out the admin-specified hostname if the IP
115     matches.
116
117 == 1.0.0_rc1 ==
118
119 * Fixes: Unique name for SE Control protocol; VO view of SA now
120   matches SA unique ID.
121 * Bugfix: Don't assume OSG_GIP_SIMPLIFIED_PATH exists
122
Note: See TracBrowser for help on using the browser.