We do this without containers quite successfully. As precaution, we... - have dedicated disk per instance. - have lots of network bandwidth, but also throttle throughout defaults. Also monitor network closely - share CPU completely. Limit Cassandra settings to limit CPU use (concurrent threads, compaction throughput, etc) and monitor closely. - have plenty of memory on top of jvm allocation - never have 2 nodes from the same cluster on a single node. - use VIPs so each instance gets its own IP address. - use sane OS defaults as documented by Amy Tobey.
That said, using cgroups or containers would provide better isolation (but worse bursting) when available. On Fri, Sep 20, 2019, 8:42 PM Sandeep Nethi <nethisande...@gmail.com> wrote: > Hi Nitan, > > You shouldn’t have any issues if you setup things properly. > > Few possible issues could be (can become a bottleneck) > > * CPU allocation (Instances can compete) > * Disk throughput & IOPS & > * Port allocations > * Network throughout > * Consistency issues. > > And we have work around for all above, > > * CPU: Use jvm file to limit number of CPU cores for each instance. > * DISK: If possible allocate dedicated disks for each instance. > * NETWORK & Ports: Have a secondary NIC (or equivalent to num of > instances). This will give you the flexibility to have Cassandra on same > ports with better networking operations. > * RACK: having multiple instances on one node can lead to consistency > problems when a hosted node goes down for some reason with having RACK’s > defined. So, this is very important to choose when going with this kind of > setup. > > Hope this helps! > > Thanks, > Sandeep.Nethi > > > > On Sat, 21 Sep 2019 at 12:20 PM, Nitan Kainth <nitankai...@gmail.com> > wrote: > >> I am looking for possible issues doing this setup without containers. >> >> >> Regards, >> >> Nitan >> >> Cell: 510 449 9629 >> >> On Sep 20, 2019, at 5:22 PM, Elliott Sims <elli...@backblaze.com> wrote: >> >> A container of some sort gives you better isolation and less risk of a >> mistake that could cause the instances to conflict in some way. Might be >> better for balancing resources between them as well, though using cgroups >> directly can also accomplish that. >> >> On Fri, Sep 20, 2019, 8:27 AM Nitan Kainth <nitankai...@gmail.com> wrote: >> >>> Hi There, >>> >>> Any feedback pros/cons for having multiple instances of C* on the same >>> machine without Docker/container solution? >>> >>> The plan is to change the ports and run multiple C* processes, so we can >>> isolate two applications as two different clusters. >>> >>
