Cassandra Pit-windows Platform compression strategy

In mid-May 2016, the cassandra3.4 version was used without a lot of tests due to the urgency of the project's launch, when it was deployed on 3 16G Windows system servers. A few months of use, most of the problems are the downtime caused by oom. In particular, there was an outage, and after restarting the database, it was discovered that memory was rising, and that was the case with multiple reboots. The possible reason for this observation is that the default compact policy taken by the database is-sizetieredcompactionstrategy (the small DB file on the disk is synthesized into a large db file). After several months of data warehousing, the strategy has produced more than 3g of DB files, do not know whether the database is a bug or memory is too small, then the compression caused by the memory leak problem. For Cassandra compression policies and upgrades, etc. are updated successively.

Cassandra Pit-windows Platform compression strategy