Advanced Configurations

we provide some configuration items for LevelDB and gRPC in config.conf file, for fine-grained performance tuning.
You may custom these items only if you have deep understanding on them, otherwise keep them as default.

LevelDB

You can custom LevelDB options in the storage part of config.conf, which looks like:

storage {
  # Directory for storing persistent data

  db.directory = "database",
  index.directory = "index",

  # You can custom these 14 databases' configs:

  # account, account-index, asset-issue, block, block-index,
  # block_KDB, peers, properties, recent-block, trans,
  # utxo, votes, witness, witness_schedule.

  # Otherwise, db configs will remain default and data will be stored in
  # the path of "output-directory" or which is set by "-d" ("--output-directory").

  # Attention: name is a required field that must be set !!!
  properties = [
    {
      name = "account",
      path = "/path/to/account",   // relative or absolute path
      createIfMissing = true,
      paranoidChecks = true,
      verifyChecksums = true,
      compressionType = 1,        // 0 - no compression,  1 - compressed with snappy
      blockSize = 4096,           // 4  KB =         4 * 1024 B
      writeBufferSize = 10485760, // 10 MB = 10 * 1024 * 1024 B
      cacheSize = 10485760,       // 10 MB = 10 * 1024 * 1024 B
      maxOpenFiles = 100
    }
  ]

}

As shown in the example above, the data of database account will be stored in the path of /path/to/account/database while the index be stored in /path/to/account/index. And, the example also shows our default value of LevelDB options from createIfMissing to maxOpenFiles. You can just refer to the docs of LevelDB to figure out details of these options.

gRPC

You can custom gPRC options in the node.rpc part of config.conf, which looks like:

node {
  rpc {
    port = 50051

    # Number of gRPC thread, default availableProcessors / 2
    # thread = 16

    # The maximum number of concurrent calls permitted for each incoming connection
    # maxConcurrentCallsPerConnection =

    # The HTTP/2 flow control window, default 1MB
    # flowControlWindow =

    # Connection being idle for longer than which will be gracefully terminated
    maxConnectionIdleInMillis = 60000

    # Connection lasting longer than which will be gracefully terminated
    # maxConnectionAgeInMillis =

    # The maximum message size allowed to be received on the server, default 4MB
    # maxMessageSize =

    # The maximum size of header list allowed to be received, default 8192
    # maxHeaderListSize =
  }
}

backup

You can custom backup options in the node.backup part of config.conf, which looks like:

node.backup {
    # my priority, each member should use different priority
    priority = 
    # members should use same port
    port = 
    # peer's ip list, can't contain mine
    members = []
}
policy: 1. the one which synchronized first will become master. 2. if synchronization is completed at the same time, the one which with big priority will become master.

E.g. create backups for node A(192.168.0.100) and node B(192.168.0.100 ): node A's configuration:

node.backup {
    priority = 8 
    port = 10001
    members = [
        "192.168.0.101"
    ]
}
node B's configuration:
node.backup {
    priority = 5
    port = 10001
    members = [
        "192.168.0.100"
    ]
}

You may refer to the source code of io.grpc.netty.NettyServerBuilder class to see details or just make a decision according to the brief comments above.