Tags: AntidoteDB/antidote
Tags
Improve DC and InterDC Functionality (#408) * Test case for issue #400 * Removed stable riak metadata and improved create dc. * Changed level of pending changes log messages * Added a better named function to add nodes to a current dc * Changed description and test case for dc descriptors. Fixed #400 * Reverted handle call for periodic updates, not needed * Bumped to version 0.2.2 * Reduced severity of log message for missing external_descriptors * Update include/antidote.hrl * Added leave_dc functionality We maintain our own dc id based on the riak_core_ring:cluster_name/1 command. Riak maintains the cluster name itself, and contrary to the comment in dc_utilities, it is restored on fail and restart correctly, even if all nodes fail at the same time. This caused errors in maintaining the ring members after simplifying the create_dc function and errors in shard count meta data. Switching back to the riak internal meta data (which is managed and updated by riak_core anyway) fixes these issues. The create_dc simplification also improved joining an empty cluster significantly (10 seconds for 8 shards). This is also work towards being able to add/remove nodes dynamically to scale after some time elapsed. Co-Authored-By: Peter Zeller <[email protected]>
Cluster management pb fix (#372) * downgrade lager because of logging problems * change test setup - forking ct_slave because it cannot be configured as I want it to - now sets the working directory for slaves, so lager logs to the correct directory - standard output of slaves is written to the ct-log and to stio.log file, which should make debugging easier - protocol buffer process now also reports stacktrace to client * also testing connect_to_dcs * use different node names for cluster management test * Adapted to cleanup of antidote_pc_codec * Upgraded to lint v0.1.10 which is checking more things. * Few fixes for the connection descriptors * Encoding ignore in pb_client * Updating the dependencies