As far as I know nodes are identified by ip address and ports defined in
yaml. If you carefully prepare configuraton you should be able to run
multiple Cassandra processes on single box participating in different
clusters. If you decide to use separate ip address and default ports it
should be quit
you have 86400 seconds a day so 42T could take less than 12 hours on 10Gb
link
19 lut 2013 02:01, "Hiller, Dean" napisał(a):
> I thought about this more, and even with a 10Gbit network, it would take
> 40 days to bring up a replacement node if mongodb did truly have a 42T /
> node like I had hear
Can you provide output from sar command for the time period when long
GC occurred ?
Regards,
Wojciech Meler
I have similar problem with cassandra 0.8.10. After digging a while
I've found that my problem is somehow related to system page scanning
activities.
After turning on -Xloggc I've found high system cpu usage during
ParNew. After looking into sar -B I've found this:
08:00:08 AM pgpgin/s pgpgout/s
I have equally balanced, 36 node cluster, RandomPartitioner, RF=3
SimpleStrategy, SimpleSnitch, cassandra 0.8.7
I'd like to change cluster topology because:
- after some reorganization in our DC 18 nodes was moved to separate room.
- I'd like to change 9 nodes to brisk to have analytical process
