Skip to content

Releases: SmartDataProjects/dynamo-consistency

v2.10

18 Jul 17:51
720e76a

Choose a tag to compare

New Feature:

  • Allow sites to be selected based on lock required. This lets gfal sites run in a separate running queue.

Bug Fixes:

  • Fix the PyPI description to not use references.
  • Fix help message in parser
  • Use exceptions when unmerged cleaning isn't working instead of exiting program
  • Make list of unlisted directories

v2.9

10 Jun 15:17
6e2bff4

Choose a tag to compare

New features:

  • Add ability to delete old files from arbitrary directories (when trying /store/logs at CERN)
  • Test with dynamo in Jenkins container
  • Add information to documentation, uses pdflatex to build now

Bug fixes:

  • Move files used in Jenkins into container
  • Fix some modules that weren't logging properly
  • Fixed dynamo.v2 so that it works in local instance

v2.8

13 Mar 17:42
735e433

Choose a tag to compare

New Feature:

  • Add check-phedex executable to double check orphan files against protected datasets on PhEDEx

v2.7

29 Jan 21:27
7276d55

Choose a tag to compare

New Features:

  • Can avoid deleting orphans through --no-orphan flag or false DeleteOrphans configuration parameter
  • Get RAL dump working. Still needs timing on operation side.
  • Allows summary table to have forced update (even when reading from cached files, for example)
  • Run over IgnoreDirectories during consistency listing (not unmerged) so that they can still be recovered when missing. Still filtered out of orphans

Bugfixes:

  • Allow previous missing files list to be empty
  • Explicitly acquire lock before accessing Sqlite3 databases to avoid crashing
  • Allow duplicate empty directory entries into history (don't know why FNAL gets this)
  • If redirector list exists for a site, use it even if the global redirector doesn't work for that site
  • A bunch of emtpy to empty, but not all of them... so that production doesn't break.

v2.6

06 Nov 16:13
ba0cbc2

Choose a tag to compare

New Feature:

  • --cnf parameter added to determine how to connect to a database using --v1 option

Bugfixes:

  • Invalid files only added to history database after redundant times missing
  • The loggers from dynamo_consistency.cms.unmerged and the loggers instantiated with the Lister objects now properly write to the output files.
  • Fixed a bug that prevented loading from cached trees in main module
  • Fixed name of dynamo_consistency.cms.filters logger

v2.5

12 Oct 18:08
b6128e3

Choose a tag to compare

New Features:

  • Added script that can dump the inventory tree or the remote tree to disk
  • Added "remote listing" method that reads file dumps done by RAL
  • Site patterns can now be matched with Python regular expressions from command line
  • Tests are now done for sl6 and sl7 (Python 2.6 and Python 2.7) on Jenkins node
  • Add page documenting CMS-specific modules

Bugfix:

  • dynamo_consistency init now fills dynamo_consistency.opts with members that may be needed to determine backend
  • Fixed some of the dynamo_consistecy.cms sub-modules' docstrings
  • Test consistency_config.json no longer checks for free memory (which killed some tests on small laptops)
  • Fixed a bug where dynamo_consistency.dynamo submodules were logging twice

v2.4

02 Oct 17:53
60f029d

Choose a tag to compare

New Feature:

  • Optionally ignore SAM tests when choosing a site to run on
  • Jenkins tests inside docker container

Bugfix:

  • Invalidation updates the persistent store properly now

v2.3

24 Sep 18:59
9b22f99

Choose a tag to compare

  • Updated documentation
  • Using history database inside of dynamo-consistency

v2.2

06 Sep 18:05
1260b43

Choose a tag to compare

  • Uses threading instead of multiprocessing
  • Jobs can be controlled via the summary database
  • There are multiple working backends for reading from dynamo

v2.1

22 Aug 14:12
5516913

Choose a tag to compare

This version implements the interface for running dynamo inside of the server process. This script still only works outside of the server though because it spawns processes. A move to threading is already under way for v2.2, so this version is ready for tagging.