On Mon, Jul 20, 2015 at 6:32 AM, Pierre Devops <pierredev...@gmail.com> wrote:
> Some of my partitions may have ~20 millions of rows, while others may have > only a few hundreds of rows. It may grow up to 300 millions of rows per > partition in a near futur, leading to very huge partition (50~80 GB I > think). > Rows this large are strongly suggestive of improper schema design. Yes, Cassandra partitions "can" contain up to 2 Billion cells. No, you probably shouldn't get anywhere near that in normal operation. > I create my sstable manually when I have a big number of rows, so I don't > go through commitlog/memtables. > > I want to know what kind of issue I may encounter with big partition > containing such a number of rows ? because now, I don't see problem on the > database when I insert or query these big partitions. However, I didn't > last it run a long time so may be problem may occurs during compaction ? > Certain operations are possible on Really Big Rows. Compaction is heavyweight but probably won't OOM. Requesting the whole row probably will probably OOM... =Rob