Depends on whether or not you’re seeing overlaps (do you have read repairs messing up the windows? Do you expire cleanly? Do you do non-ttl deletes?)
When I wrote it I used to run about 70-80% full, but we were willing to do some manual cleanup when overlaps prevented dropping full sstables -- Jeff Jirsa > On Jul 17, 2018, at 5:35 PM, Eunsu Kim <eunsu.bil...@gmail.com> wrote: > > Can I ask you an additional question here? > > How much free space should I have if most tables use > TimeWindowCompactionStrategy? > > >> On 13 Jul 2018, at 10:09 PM, Vitaliy Semochkin <vitaliy...@gmail.com> wrote: >> >> Jeff, thank you very much for reply. >> Will try to use 4TB per instance. >> >> If I understand it correctly level compaction can lead to 50% >> https://docs.datastax.com/en/dse-planning/doc/planning/planningHardware.html >> >> Regarding the question of running multiple instances per server, am I >> correct that in case of 3.11 instances and having several disks >> dedicated for each instance, running multiple instances per server is >> ok? >> >> >>> On Thu, Jul 12, 2018 at 5:47 PM Jeff Jirsa <jji...@gmail.com> wrote: >>> >>> You can certainly go higher than a terabyte - 4 or so is common, Ive heard >>> of people doing up to 12 tb with the awareness that time to replace scales >>> with size on disk, so a very large host will take longer to rebuild than a >>> small host >>> >>> The 50% free guidance only applies to size tiered compaction, and given >>> your throughput you may prefer leveled compaction anyway. With leveled you >>> should target 30% free for compaction and repair >>> >>> You don’t need more than one Cassandra instance per host for 4tb but you >>> may want to consider it for more than that - multiple instances are >>> especially useful if you have multiple (lots of) disks and are running >>> Cassandra before CASSANDRA-6696 (which made jbod safer). >>> >>> -- >>> Jeff Jirsa >>> >>> >>>> On Jul 12, 2018, at 7:37 AM, Vitaliy Semochkin <vitaliy...@gmail.com> >>>> wrote: >>>> >>>> Hi, >>>> >>>> Which amount of data Cassandra 3 server in a cluster can serve at max? >>>> The documentation says it is only 1TB. >>>> If the load is not high (only about 100 requests per second with 1kb >>>> of data each) is it safe to go above 1TB size (let's say 5TB per >>>> server)? >>>> What will be safe maximum disk size a server in such cluster can serve? >>>> >>>> Documentation also says that compaction requires to have %50 of disk >>>> occupied space. In case I don't have update operations (only insert) >>>> do I need that much extra space for compaction? >>>> >>>> In articles (outside Datastax docs) I read that it is a common >>>> practice to launch more than one Cassandra server on one physical >>>> server in order to be able use more than 1TB of hard driver per >>>> server, is it recommended? >>>> >>>> --------------------------------------------------------------------- >>>> To unsubscribe, e-mail: user-unsubscr...@cassandra.apache.org >>>> For additional commands, e-mail: user-h...@cassandra.apache.org >>>> >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: user-unsubscr...@cassandra.apache.org >>> For additional commands, e-mail: user-h...@cassandra.apache.org >>> >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: user-unsubscr...@cassandra.apache.org >> For additional commands, e-mail: user-h...@cassandra.apache.org >> >
