Antidote configuration

[albsch]

Unified and simplified Antidote configuration.

• Changed shell command to use the release shell instead
• Added part on how to use the relase in the Readme
• Rebar now understands symlinks for _checkouts, fixed readme
• Removed deprecated env entrypoint script
• Removed launch-nodes script, uses legacy RELX_REPLACE_OS_VARS functionality
• Removed all schema files except the riak_core schema for compatibility purposes. Especially the erlang-vm.schema file caused problems when trying to use the NODE_NAME env variable
• Configuration uses the new .src functionality to enable native OS variable replacement in the root configuration file
• Split up smaller configuration files, currently only for the network configuration
• Enabled packing files (data antidote, data riak, logging files) in a root directory if specified via ROOT_DIR_PREFIX
• Same for logging and data dirs (LOGGER_DIR_PREFIX, DATA_DIR_PREFIX)
• Removed vars.config, it was only used for the Mustache templating system of Rebar, which was only used for the schema files of cuttlefish, which we do not use
• Fixed ranch to version 1.7.1 for now, otherwise it would upgrade to a release candidate with breaking API changes
• Bumped antidote to version 0.2.1 (both rebar.config and antidote.app.src)
• Moved documentation from antidote.app.src to config/sys.config where all configuration for the release resides
• Added a validation function to ensure the antidote data directory exists
• Commented out stats calls which are not implemented yet in the antidote_stats module (metadata_updated, metadata_update_stable)
• Removed partially obsolete max ETS tables argument. The amount is not limited anymore, and 8000 is a reasonable default size of the internally used table for named tables.
• Removed the ERL_MAX_PORTS env parameter. This is no longer used and is replaced by vm argument +Q. Furthermore, the default is much higher (65536) than this value (4096).
• Added kernel network config for documentation purposes
• Removed eunit_opts
  • Verbose output is worse than default output, as it doesn't show failures at the end
  • Eunit_logs is not useful
• Added eunit.coverdata to coverage stats, coverage has increased by ~7%
• Moved monitoring setup to antidote_stats repository
• Fixed logging output path in systests and removed copied erlang code for ct_slave_ext
• Redirected lager output to logger via custom lager backend

Consequences:

• Update readme for monitoring setup in antidote_stats repository
• Update documentation for monitoring setup in Antidote documentation
• Docker files have to be updated
• Update documentation for Antidote on how to use a release, go into more detail than the readme
• Some configuration parameters cannot be directly configured by OS variables and have to be configured dynamically via the -Application Key Value parameter tuple, for which Value can be used for OS variable configuration
• VM arguments are copy pasted from a rebar default from many years ago. Check if vm args improve performance, if not disable them
• Check if there is a way to specify coverage output path for ct_run

[albsch]

The new dockerfile can be seen here.

[bieniusa]

Why are we bumping from v0.1.1. to v0.4.0?

[albsch]

Why are we bumping from v0.1.1. to v0.4.0?

There are multiple versions everywhere and I didn't know which version to take. The docker-antidote version is later than v0.2.0 and the release configuration (rebar) is 0.0.2. So I guess the version should be something like v0.2.1 everywhere?

Actually, I don't know how the rebar release version and the app.src version relates to the release tagged version/dockerfile version. Are they related at all?

[albsch]

New statistics: