forked from holzman/gip
-
Notifications
You must be signed in to change notification settings - Fork 0
/
changelog
320 lines (274 loc) · 13 KB
/
changelog
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
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
Changelog
== 1.3.11 ==
* SLURM: Change GlueCEPriorityPriority from 0.0 to 0 (it needs to be an int)
* Add patch to read Info Services section from OSG config file if present (SOFTWARE-1382)
* Change default BDII servers for ITB sites (SOFTWARE-1406)
* Read user-customized .rvf files in /etc/globus/gram if they are present (SOFTWARE-1527)
* Silence DeprecationWarnings (SOFTWARE-1536)
== 1.3.10 ==
* Add max_wall option for PBS/LSF/Slurm
* PBS bugfix: if the last queue in qstat doesn't have the 'started' or 'enabled'
attribute, GIP would crash. (We fixed the other queues in 1.1.8)
== 1.3.9 ==
* Official support for the slurm batch system
* New parameter for PBS/LSF/Condor/Slurm: idle_slack
For VOViews, setting idle_slack in the relevant batch system
section of the config.ini will affect how free slots are
reported in the VOView according to:
if idleJobs > idle_slack: freeSlots = 0
== 1.3.8 ==
* Make child reads non-blocking (contributed by John Brunelle @ Harvard)
== 1.3.7 ==
* Correct search path for RVF files
* Remote gip_diagnostic can parse *-attributes.conf
== 1.3.6 ==
* Prepend rather than append GIP_LOCATION python paths to sys.path
(if defined)
* Removed "Using GIP SVN revision" line from logs
* Added Condor-paths where they might be missing
* Condor: parse hierarchical groups correctly
== 1.3.5 ==
* Ignore condor_config in .ini when set to DEFAULT, UNAVAILABLE, etc.
* Update osg-version location to /usr/bin by default
== 1.3.5rc1 ==
* Check for duplicate key/values (ignoring case) -- RFC4512 forbids it
* Ignore comment lines in read_ldap
* Remove GlueLocation stanzas for software publication
* Remove configure_gip placeholder
== 1.3.4 ==
* Better error messages for invalid vo maps
* GIP will now look in the config files for the proper user for dropping privileges
== 1.3.3 ==
* Fix wrong location for gip log (OSG SOFTWARE-374)
* Drop privs to user tomcat instead of user daemon
* Publish GRAM version as 5.0 instead of 2.0
* PBS: only add queued jobs as waiting (were counting all non-running jobs
as queued before)
== 1.3.2 ==
* Fix gip_info bug (OSG SOFTWARE-290)
* Clean up gip-validator (MYOSG-25)
* Forward port non-blocking read code (it didn't make it from 1.2.x to 1.3)
== 1.3.1 ==
* Bug fix -- we no longer exit with an error if we don't find config.ini
== 1.3.0 ==
* Update for RPM install - following FHS, Fedora, and OSG guidelines
== 1.2.0alpha0 ==
* Suppress publishing GlueVOInfoTag unless there is a space reservation
* Implement CREAM support
* Implement HTPC support
* PBS: Better calculate free slots for queues
* Condor: Bug fix for exclude_schedds
* Condor: Support query_only_local_condor option
* LSF: Bug fix (we were publishing LSF queues _only_ if they were _not_ in
the RVF, which is the exact opposite to what we should do).
* Remove spaces in storage AccessControlBaseRule
* Remove duplicate software locations
* Tighter exception handling for parsing config files, imlpying we should:
* Explicitly return the default for blank config section or option
* Avoid executing temporary/backup files
* Updated unit tests
== 1.1.14 ==
* Bugfix: fix glexec support publishing
== 1.1.13 ==
* Bugfix: fix the ACBRs to VOLocalViews
* Bugfix: enable allowed_vos to also work properly on the storage element
* LSF: better parsing of waiting jobs, free slots, host groups
* LSF: bugfix: don't include queues that are not in the RVF
* PBS: better chance to include PBS binaries in the path
* SGE: fix missing includes
* Enable publication of WLCG site attributes (admins must set wlcg_tier,
wlcg_parent, wlcg_name in the [site] section of config.ini)
* Better reporting of commands with non-zero exit status
* Remove duplicates from PATH
* Updated unit tests
== 1.1.12 ==
* gip-validate syntax fixes
* Implement __ne__ for ldap
== 1.1.11 ==
* Support publication of arbitrary SE port numbers via srm_port configuration
parameter (even though they are deprecated in the schema)
* Children with non-zero exit status are now logged properly
* Added support for local template dirs (contributed by Sam Morrison @ ARCS),
needs local_template_dirs parameter to be defined
* Added new tools: gip_diagnostic, gip-validator.py
* Bugfix: continue if we can't open gsi-authz.conf
* Condor: Allow exclusion of 'default' queue via excluded_groups parameter
* Condor: Remove unpublished CEs from foreign keys. The CEs were removed
from publication because they did not match to any VOs.
* LSF: Bootstrap LSF by sourcing the lsf_profile parameter (if not provided,
we try /lsf/conf/profile.lsf)
* LSF: Ensure max wall time is an integer
* LSF: Use -w option in LSF commands to avoid truncation
== 1.1.10alpha0 ==
* Publication of glexec (if glexec_location is published in the Misc Services
section of the config.ini)
* Allow and document standalone SE publication (if se_only = True in the GIP
section of the config.ini)
* Revised option names (excluded_groups -> exclude_groups,
exclude_schedd -> exclude_schedds)
* Condor: if the sum of the group quotas are larger than the size of the
cluster, set the TotalCPUs to the size of the cluster (previous behavior was
using 999999)
* Condor: Fix VOView case-sensitivity bug
* Condor: count flocked jobs as running by default
* Condor: Set default group name to 'default' if negotiator isn't aware of it.
== 1.1.9rc1 ==
* PBS: Properly drop PBS queues that are not in the RVF.
* Condor: Add capability to ignore condor accounting groups
* Condor: Add capability to query collector instead of negotiator
* Condor: Add capability to exclude schedds during condor_status -submitter
* Change minimum EstimatedResponseTime to 60
== 1.1.8 ==
* Don't drop misconfigured PBS queues (that lack enabled or started attributes)
* Get siteName correctly from config.ini (first try resource_group, then
site_name, then resource, and as a last resort, resource_group_name)
* dCache: advertise the endpoint as srm_endpoint in config.ini (if set)
* dCache: do better parsing link groups and pool groups
* dCache: summarize total SE information correctly
== 1.1.7 ==
* Addition of resource_group_name
* Set free slots = 0 if jobs are waiting
* check for jobs_constraint == 'True' is now case-insensitive
* condor_q -xml parsing understands both old (pre 7.3.2) and new format
== 1.1.6 ==
* Add GlueLocationLocalID OSG_VERSION to publish OSG version in friendlier way
* Add cpu_platform publication capability
* Add -SITENAME to the end of Subcluster Unique IDs in an attempt to make them globally unique
* Improved error message on logfile permission problems
* SL3 bugfix for try/except/finally block
* Update unit tests
== 1.1.5 ==
* Help for debugging: print more python backtraces in logs
* Add $PBS_LOCATION/bin to pbs path (for PBS/torque)
* Code that renames VOs now matches only uscms* and usatlas* instead of us*
* Change GlueHostOperatingSystem{Name,Version,Release} to conform to EGEE standard (http://goc.grid.sinica.edu.tw/gocwiki/How_to_publish_the_OS_name)
* Publish $OSG_GRID as new GlueLocationName with new osg_info_grid_dirs.py provider
* Force provider children to always, always exit
* Set provider log to be module.log instead of module.logs
* Remove unused libexec/osg_info_provider_cluster.py
* Remove unused code from providers/cluster.py
== 1.1.4 ==
* Ability to advertise local mount points
* Publish HEPSPEC if configured
* If uid=0, drop privileges to daemon (will help prevent root from clobbering
daemon-owned files)
* When vdt-version fails, publish "OLD_VDT"
* Remove some deprecated Glue attributes
* More debugging information when srm-ping fails
* Many SGE bug fixes (thanks, Iwona!)
* New sge_cell parameter in the [sge] section of config.ini; without it, we presume the SGE
commands work are found in the environment and work.
* New other_ces parameter to publish other CEs in the cluster (for the CE to
SE bindings)
* Enforce condor invariants -- we should do a better job reporting the number
of free and running slots
== 1.1.3 ==
* If storage_vos is set, ignore the blacklist/whitelist for storage (storage_vos is an explicit whitelist already).
* Fix bug where we tried to remove duplicate stanzas more than once
* Insert newlines before stanzas if needed
* RHEL3 compatibility: fix integer overflows in dCache 1.9 provider
* RHEL3 compatibility: fix SE-related config bug
* Check for osg-version in both $VDT_LOCATION and $VDT_LOCATION/osg/bin
* Condor: Allow (groupname)_blacklist and (groupname)_whitelist in Condor
* Condor: Parse condor_status correctly for multi-schedd sites
* PBS: enforce job slot invariants:
max_total <= max_running
free_slots <= max_running
assigned <= max_running
== 1.1.2 ==
* Include both old (voname) and new (VO:voname) ABCRs in GlueSA ACBRs
* Now correctly identify the default SE to eliminate false "UNKNOWN" values for the Classic SE
* Fixed setting clusters and cesebind in config_ini and blank acbrs in the SA
* Fix to allow multiple SE's for a single CE
* Added new unittests for CESEBind and Cluster overides
* Unittests now will run without any outside configuration dependencies
== 1.1.1 ==
* allow old-style storage config again (GIP stanza)
* BeStMan: correctly determine SE space when no space tokens exist
* BeStMan: improved logging
* Parse updated srm-ping output
* Fix crashing LSF provider
== 1.1.0rc1 ==
* More fleshed out (but probably still broken) SGE provider
* Fix to include default SE and publish correctly
* Revert to config.ini instead of current-config.ini
== 1.1.0 ==
* Removal of gip-attributes.conf and configure_gip perl script, because:
* We now read config.ini directly
* Reconfigured GIP providers to determine the correct batch system and storage
system on-the-fly, removing the need for admin configuration
* We do not attempt to auto-detect subclusters
* New wrapper script with new name (osg_info_wrapper.py): (CEMon
compatibility issue)
* Wrapper script supports a new delete-attributes.conf
* Completely new providers for Condor, LSF, and SGE. SGE and LSF are lightly
tested (SGE probably does not yet work out of the box).
* Condor provider now advertises data consistently with the other batch
system providers.
* New BestMan provider
* New dCache provider
* New provider for dCache 1.9 which does NOT need the dCache admin interface
* Improved adherence to Glue 1.3
* Increased unit test coverage
* Glue parameters CESEbind, site, subclusters, and cluster entries are all
dynamic
* Improved and more secure method for advertising VO-installed software
* Glue 1.3 style AccessBaseControlRules are used (CEMon compatibility issue)
* Storage and subcluster configurations via config.ini have been refactored
to be simpler and more straightforward.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
== 1.0.8 ==
* Fix missing and extraneous imports in token_info_provider.py
== 1.0.7 ==
* In configure_gip, check for globus only when looking for LSF queues
== 1.0.6 ==
* Include plugin to calculate GlueCEStateEstimatedResponseTime ("ERT")
* Include logging for ERT
* Allow attributes in the DN to have internal whitespace
== 1.0.5 ==
* Fix syntax error (spurious backticks) in configure_gip
== 1.0.4 ==
* Bugfix for GlueVOInfoLocalID, which needs to include GlueSEUniqueID.
== 1.0.3 ==
* Sets reasonable defaults for GlueCEPolicyMaxObtainableWallClockTime, GlueCEPolicyMaxObtainableCPUTime, GlueCEStateWorstResponseTime
== 1.0.2 ==
Bugs fixed:
* Respect max_queueable for PBS (10)
* Multiple SE per CESE bindgroup (11)
* Schema compliance for GlueServiceStatusInfo?, GlueSARetentionPolicy,
GlueSAAccessLatency, GlueSAExpirationMode,GlueSACapability,
GlueSEImplementationName,GlueSEStatus (9,12)
* Custom attributes via alter-attributes (14)
* PBS Default wallclock set to 6000 instead of 0
Features added:
* Logging (13)
* LSF queues parsed from RVF instead of bqueues (thanks Alan Sill)
== 1.0.1 ==
Bugs fixed (references to GIP ticket numbers in [])
- The dynamic providers were parsing the VO list differently than in
previous cases [1]
- For an SRM SE, GlueSEArchitecture is changed from "disk" to "other".
This avoids the classicSE script trying to execute df in the wrong
places [2, 8]
- $GIP_LOCATION/etc/gip.conf was not being read in some cases due to
invalid path expansion [3]
- Multiple collector hosts were not getting parsed correctly (GIP
ticket [4]
- OSG Version is now published along with the software tags [5]
- PBS dynamic provider was not parsing waiting jobs correctly (GIP
ticket [6]
- For non-PBS sites using dynamic dcache, GlueSEUniqueID was not being
correctly generated [7]
- GlueCESEBindCEAccessPoint will not double-count the GlueSAPath now
- Wrong permissions and ownership were generated for
dcache_storage.conf, gip.conf
- CESEBind now advertises /UNAVAILABLE instead of dropping the
attribute if the access point is not set
- SRM hostname was being advertised based on reverse DNS, which is a
problem for sites where the SRM host is a CNAME instead of an A
name. We now print out the admin-specified hostname if the IP
matches.
== 1.0.0_rc1 ==
* Fixes: Unique name for SE Control protocol; VO view of SA now
matches SA unique ID.
* Bugfix: Don't assume OSG_GIP_SIMPLIFIED_PATH exists