Disable compaction and gc grace on cassandra

I always insert PRIMARY KEY data ((site_name, date), time, id), while the site_name and date can be the same, the time which is the tamed field and id (uuid) is different. Therefore, I always add new data. Data is inserted with TTL (currently 3 days). Since I am not deleting or updating, can I disable the compaction? There is a TTL review. Will it affect something? Also, since no entry has been deleted, can I turn off the gc_grace time? I want to put as little load on the servers as possible. Really appreciate if anyone can help?

+4
source share
2 answers

TTL create tombstones. Thus, a seal is required. If your data is time series data, you can consider a new date compilation: http://www.datastax.com/dev/blog/datetieredcompactionstrategy .

If you use TTL and set the grace to 0, you are asking for problems if your cluster is not the only node. Grace is the time to wait before collecting tombstones. If it is 0, it will not wait. This may seem good, but in fact it means that โ€œdeletionโ€ may not spread throughout the cluster, and deleted data may reappear (for example, other nodes may have it, and the last current value will โ€œwinโ€). This data type called zombie data. Zombies are bad. Do not feed the zombies.

: http://www.datastax.com/documentation/cassandra/2.1/cassandra/tools/toolsDisableAutoCompaction.html. , , . , .

, . , .

+3

Source: https://habr.com/ru/post/1570328/


All Articles