API for Runnables written in "guest languages" (with reference Python implementation and examples)
Test suite (inspired by Roy's original pull request)
"TailTrimmer" [ in analyses with nontrivial batch sizes ] several techniques are now used to automatically decrease the batch size
towards the end of the analysis in order to speed up the execution of the whole analysis
Stability improvements that significantly increase efficiency of parallel execution
[ higher level features ]
* support for Runnables written in Python3 and API for extending similar support to other languages (this API may still change)
* coloured Beekeeper output - catches the eye!
* SystemCmd now runs through Capture::Tiny , captures the error output from the actual command that gets stored in log_message
* SystemCmd also knows how to capture MEMLIMIT events from the underlying Java code
* SystemCmd can map specific return codes to dataflow events
* a new DbCmd runnable that mimics the behaviour of db_cmd.pl script ; you can also pipe data in or out of the connection to another system command
* DbCmd, DatabaseDumper and MySQLTransfer runnable hide passwords in the command lines that they run
* beekeeper.pl -unkwn option to clean up the workers found to be in UNKWN state (at the user's risk!)
[ lower level features ]
* record the 'meadow_user' in each Worker entry -- these values are also used when querying the Meadow to avoid running an equivalent of "-u all" in SGE Meadow
* record the 'when_seen' timestamp in each Worker entry -- when the Worker was last seen as running by the Beekeeper process.
* testing: introduced a Travis-integrated test suite loosely based on Roy's original pull request
-- the extended version tests direct API calls, runs individual Runnables (and tests their dataflow/warning events) or whole pipelines
* testing: Travis runs tests against Hive databases stored in local MySQL, PostgreSQL and SQLite databases
* stability [too many simultaneous queries] : detect and log deadlock collisions and retry them for a given number of times before failing
* stability [running out of server connections] : try to resolve the "too many connections" situation by bouncing, waiting and retrying
* stability [running out of local ports] : avoiding RELOCATED workers by applying incemental backoff-and-retry approach from Ethernet CSMA/CD protocol
* stability [applying an incorrect patch] : schema patches now have internal SQL-based checks and should not cause much damage if applied in wrong order
+ a new script to create such patches
* the schema version changes to 73
* multiple bug fixes, many of which have been ported to the previous version branches.
