Sorry for spamming, folks. Okay, I found out why it is growing.
system.paxos row count is actually 3% greater than the data table. It may be caused by re-insertion of some data range. My ultimate question now would be why the paxos rows are not getting removed after the transaction. There is high disk IO on the directory of the server where there are paxos table files. And the insertion gets slower and slower. (Yes, it seems to be somewhat anti-pattern since we are using 'IF NOT EXISTS' massively, yet, still, I don't understand why it is getting slower like this.) Thanks. Haebin 2014-06-03 11:20 GMT+09:00 Frederick Haebin Na <haebin...@gmail.com>: > Sorry, I found out what are in the system.paxos. > > select * from system.paxos limit 1; > > | row_key | cf_id | in_porgress_ballot | most_recent_commit | > most_recent_commit_at | proposal | proposal_ballot | > > I will investigate more on why it is growing continuously. > If anyone have an idea why it is like this, please, share your experiences. > > Thank you. > > Haebin > > > > 2014-06-03 11:00 GMT+09:00 Frederick Haebin Na <haebin...@gmail.com>: > > Hello all, >> >> We are trying to migrate data with 'INSERT IF NOT EXISTS' clause. >> Yet, strange thing is that the system.paxos table gets larger like 100GB >> which is equal to the size of the data table. >> >> Does anyone know what is happening here? >> What does system.paxos table store? >> >> Thank you. >> >> Haebin >> > >
