* Port agency performance tuning for many shards to devel.
* Add more IDs to LOG_TOPIC calls.
* Even more IDs for LOG_TOPIC.
* Fix a duplicate LOG_TOPIC ID.
* Fix an old merging bug in devel.
* Don't hesitate between phases one and two for small clusters.
* precondition plan / version in compaction / store TTL removal independent of local _ttl set
* Agency init loops break when shutting down.
* assertion failures in store on restarting following agents
* Minor porting fixes from 3.4
* Ignore satellite collections in shrinkCluster in agency.
* Abort RemoveFollower job if not enough in-sync followers or leader failure.
* Break quick wait loop in supervision if leadership is lost.
* In case of resigned leader, set isReady=false in clusterInventory.
* Fix catch tests.
* agents' is obtained from leader's configuration
* corrections in Supervision for advertised endpoints
* change log
* Updated Documentation for cluster/health.
* Unified naming convention.
* Fixed missing update of volatile fields.
* Set version in right order.
* Removed debug output.
* Fixed jslint - missing ;
* indexes are marked while still missing in Current
* index handling getCollection
* supervision gets indexes from isbuilding, when coordinator is gone before finishing
* seems right now
* fixed broken views
* remove junk comments
* cleanup
* node / supervision adjustements
* supervision fixes
* neunhoef remarks part i
* neunhoef remarks part ii
* neunhoef remarks part ii
* neunhoef remarks part iiI
* collection's current version please
* no need to wait for current once again
* no longer necessary code
* clear comments
* delete left overs
* dead code revived
- Schmutz now called "Maintenance" and completely implemented in C++
- Fix index locking bug in mmfiles
- Fix a bug in mmfiles with silent option and repsert
- Slightly increase supervision okperiod and graceperiod
* Initial low level interface for thread crash reporting (and management).
* Add a member version of isClusterRole()
* isolate heartbeat thread creation to new StartHeartbeatThread(). create heartbeat thread even if not a cluster or if an agent.
* update runDBServer() and runCoordinator() to shutdown more quickly by polling isStopping() at additional locations.
* copying updates from different branch / PR
* basic thread crash logging. Not yet tied into Agency arangod or have any specific threads posting crashes
* make Supervision thread a CriticalThread
* sandwich CriticalThread between Thread and other classes to create long term, repeating thread crash reporting.
* restore code lost upon branch update relating to new startHeartbeatThread() function
* add CriticalThread.cpp to build
* add new runAgentServer() function to loop for Agents. Make Heartbeat thread derive from CriticalThread.
* remove debug line
* Supervision goes to Maintenance mode, when /arango/Supervision/Maintenance exists
* coordinator route stands
* stop updates in transient, when supervision off
* safety checkin of Node throw reduction.
* final round of Node throw protection. Common accessors now protected to force code to hasAsXXX() functions.
* supervision must not work as long as agent is still preparing
* leadersince atomic and pushed to end of leader preparation
* More consistent use of integer types.
* Slightly change order of events in Supervision loop.
* let's not miss failedserver removal
* remove resetting of FailedServers in test code
* Only call abortRequestsToFailedServers at most every 3 seconds.
* Really enforce the hidden option --server.maximal-threads if given.
* Switch off --log.force-direct in scripts/startStandAloneAgency.sh
* Lower the timeout for sending AppendEntriesRPC to 150s.
* Erase _earliestPackage when becoming a leader.
* Challenge leadership in agent main loop.
* Use steady_clock for _earliestPackage.
* Change _lastAcked and _leaderSince to steady_clock as well.
* time difference calculations based on old readSystemClock to steadyClockToDouble
* All system_clock transitioned to steady_clock in Agent. Remaining system_clock are user input / output or timestamps
* Inception system_clock to steady_clock
* Server short name could arrive too late for first health check. Would lead to supervisor thread crash. Add test for this condition and defense against other unknown throws in health check.
* Correct capitalization of ShortName. Add spaces to two Log lines.
Lars Maier
Jan
Dan Larkin-York
jsteemann
Markus Pfeiffer
Frank Celler
Kaveh Vahedipour
Max Neunhöffer
Jan Christoph Uhde
Simon
Manuel Pöter
Frank Celler
Matthew Von-Maszewski