If you have one snapshot of all your sstables and then you do a major compaction, your disk space will double. So that is what is causing your disk space to go up. Not the snapshot after the major compaction.
Snapshot to NFS resource - it is possible
(22 posts) (3 voices)-
Posted 5 months ago #
-
And what about minor compactions? If I have default settings, C* will do minor compactions when reach max sstables created, so a new sstable will be created as result of this.
I'm right or I say something wrong?Posted 5 months ago # -
No, that is correct. After minor compactions cassandra will delete the old sstables that were compacted into a new sstable. If you have snapshots, deleting those files won't actually free any disk space since they still exist in the snapshots directory.
Posted 5 months ago # -
Hence, any SSTABLE change, generates a file in snapshot directory, and this action produces some sort of "nulling" hard link pointer?
Posted 5 months ago # -
You've lost me. The only thing that generates files in the snapshots directory is running the snapshot command. That creates hard links to all live sstables.
Posted 5 months ago # -
Yeah, right, I understood, but if I run daily major compactions, snapshots always change (in our situation, with tombstoned rows) and then take snapshots, every older snapshot folder finally have real files, not hard-linked pointers.
Posted 5 months ago # -
Kind of. Cassandra will have deleted the original pointer to the file, leaving only the snapshot pointer to the file if that is what you mean. I would recommend reading http://en.wikipedia.org/wiki/Hard_link
Posted 5 months ago #
