Write-ahead log

    It is used to run data recovery after a server crash, and can also be used ina replication setup when slaves need to replay the same sequence of operations ason the master.

    By default, each write-ahead logfile is 32 MiB in size. This size is configurable via theoption —wal.logfile-size.When a write-ahead logfile is full, it is set to read-only, and following operations willbe written into the next write-ahead logfile. By default, ArangoDB will reserve somespare logfiles in the background so switching logfiles should be fast. How many reservelogfiles ArangoDB will try to keep available in the background can be controlled by theconfiguration option —wal.reserve-logfiles.

    Long-running transactions prevent write-ahead logfiles from being fully garbage-collectedbecause it is unclear whether a transaction will commit or abort. Long-running transactionscan thus block the garbage-collection progress and should therefore be avoided at all costs.

    On a system that acts as a replication master, it is useful to keep a few of the already collected write-ahead logfiles so replication slaves still can fetch data fromthem if required. How many collected logfiles will be kept before they get deleted isconfigurable via the option —wal.historic-logfiles.

    RocksDB WAL Details

    The options mentioned above only apply for MMFiles. The WAL in the rocksdb storage engineworks slightly differently. Note: In rocksdb the WAL options are all prefixed with. The options do have no effect.

    The individual RocksDB WAL files are per default about 64 MiB big. The size will always be proportionallysized to the value specified via . The value specifies the amount of data to build up in memory (backed by the unsorted WAL on disk) before converting it to a sorted on-disk file.Larger values can increase performance, especially during bulk loads. Up to write buffers may be held in memory at the same time, so you may wish to adjust this parameter to control memory usage. A larger write buffer will result in a longer recovery time the next time the database is opened.