GitLab

* ensembl-github/master:
  bugfix: round down the potentially non-integer hive_capacity-related limit
  bugfix: switching to using the newly created find_all_by_pattern() and collections in Worker to fix the -job_id bug
  find_all_by_pattern() is now supported on the collection level

bugfix: switching to using the newly created find_all_by_pattern() and collections in Worker to fix the -job_id bug

Merge remote-tracking branch 'ensembl-github/master' into python

* ensembl-github/master:
  moved mysql_use_result=>1 to the $sth_attribs hash
  allow capturing dataflow from a standaloneJob into a semaphored group
  Dataflow rules are stored in the Analysis, not the Job
  bugfix: make sure dataflown jobs are stored even in standalone mode
  generalized load_parameters() moved into AnalysisJob
  bugfix [to 97e17d94] : removed the leading zero from param_init call in gc_dataflow
  The $a syntax in #expr() has long gone, Should use #a# instead
  Deprecate --append in db_cmd.pl
  deprecated and removed support for strict_hash_format() method in Runnables
  We don't use $SKIP_TESTS in ensembl-hive

Conflicts:
	modules/Bio/EnsEMBL/Hive/Worker.pm
	t/10.pipeconfig/longmult.t

* master: (73 commits)
  revised checklist
  bugfix [to ENSCOMPARASW-191] : revive the ability to create a job from (analysis_id, input_id) pair
  only show the extra details about claiming in log_message table when -debug is on
  new formula for estimating num_required_workers
  pre-load Analyses/AnalysesStats objects compatible with the Worker for faster access
  a prototype tail-trimming formula (only works for analyses where hive_capacity is defined)
  cosmetic: fixed a remark
  an ad-hoc attempt to pull up the number of required workers
  account for the remaining_job_count when computing AS::get_or_estimate_batch_size
  bugfix: cache the AS object during the batch instead of fetching for every job
  bugfix: do not claim if there are no jobs
  bugfix: make sure special_batch workers do not unclaim
  some TT-related debug messages sent to log_message table
  Workers attempt to unclaim the rest of the batch to share the load (tail trimming)
  Trim the long tails of heavily-batched analyses (and plans of how to implement UNCLAIMing)
  back-ported from version/2.3 the bugfix [to ENSCOMPARASW-247] : find-and-match the funnel job when storing it failed
  back-porting the LMA::store_hive_message method
  bugfix [to ENSCOMPARASW-247] : find-and-match the funnel job when storing it failed
  bugfix to [to bee183d1] : ignore a possible collision with another beekeeper updating worker_resource_usage
  extended LMA::store_worker_message to work with worker_id
  ...