Re: is there an upper limit of the size tier of stcs?
If you exceel 50% then you could be in trouble for compaction. No immediate harm though, just that compaction will start failing and eventually, you will see read latencies go high. On Mon, Jan 14, 2019 at 4:59 AM Jinhua Luo wrote: > But what if the sstable file size of the largest tire is too large to > compact (e.g. no spare disk room to merge two huge sstables). > Not to mention that huge file is likely to bring troubles for some > file systems, and it's hard to backup them. > > Nitan Kainth 于2019年1月14日周一 上午1:53写道: > > > > No, sstables keep compacting when they meet criteria. Look for > compaction sub properties > > > > > > Regards, > > > > Nitan > > > > Cell: 510 449 9629 > > > > > > On Jan 13, 2019, at 8:18 AM, Jinhua Luo wrote: > > > > is there a largest tier where the sstables stop merging anymore? if so, > then does it mean some tombstones at that last tier would not be removed? > if not, then it's hard to imagine to compact sstables in huge size, like > 10gb; after all, the total size of sstables would definitelly grow, so a > huge sstable would be very likely possible, especially when the range of > the keys is wide. > > - > To unsubscribe, e-mail: user-unsubscr...@cassandra.apache.org > For additional commands, e-mail: user-h...@cassandra.apache.org > >
Re: is there an upper limit of the size tier of stcs?
But what if the sstable file size of the largest tire is too large to compact (e.g. no spare disk room to merge two huge sstables). Not to mention that huge file is likely to bring troubles for some file systems, and it's hard to backup them. Nitan Kainth 于2019年1月14日周一 上午1:53写道: > > No, sstables keep compacting when they meet criteria. Look for compaction sub > properties > > > Regards, > > Nitan > > Cell: 510 449 9629 > > > On Jan 13, 2019, at 8:18 AM, Jinhua Luo wrote: > > is there a largest tier where the sstables stop merging anymore? if so, then > does it mean some tombstones at that last tier would not be removed? if not, > then it's hard to imagine to compact sstables in huge size, like 10gb; after > all, the total size of sstables would definitelly grow, so a huge sstable > would be very likely possible, especially when the range of the keys is wide. - To unsubscribe, e-mail: user-unsubscr...@cassandra.apache.org For additional commands, e-mail: user-h...@cassandra.apache.org
Re: is there an upper limit of the size tier of stcs?
No, sstables keep compacting when they meet criteria. Look for compaction sub properties Regards, Nitan Cell: 510 449 9629 > On Jan 13, 2019, at 8:18 AM, Jinhua Luo wrote: > > is there a largest tier where the sstables stop merging anymore? if so, then > does it mean some tombstones at that last tier would not be removed? if not, > then it's hard to imagine to compact sstables in huge size, like 10gb; after > all, the total size of sstables would definitelly grow, so a huge sstable > would be very likely possible, especially when the range of the keys is wide.
is there an upper limit of the size tier of stcs?
is there a largest tier where the sstables stop merging anymore? if so, then does it mean some tombstones at that last tier would not be removed? if not, then it's hard to imagine to compact sstables in huge size, like 10gb; after all, the total size of sstables would definitelly grow, so a huge sstable would be very likely possible, especially when the range of the keys is wide.