Hello. I am using Cassandra 0.6.1 on ubuntu 8.04. 3 node cluster.
I notice that when I start making lots of read requests (serially), memory
usage of jsvc keeps climbing until it uses up all memory on the server (happens
for all 3 servers in the cluster). At that point, the box starts
On Wed, Apr 28, 2010 at 12:12 PM, Kyusik Chung kyu...@discovereads.com wrote:
Hello. I am using Cassandra 0.6.1 on ubuntu 8.04. 3 node cluster.
I notice that when I start making lots of read requests (serially), memory
usage of jsvc keeps climbing until it uses up all memory on the server
Hi Ryan,
Do you mean these settings, or other settings?
SlicedBufferSizeInKB64/SlicedBufferSizeInKB
FlushDataBufferSizeInMB32/FlushDataBufferSizeInMB
FlushIndexBufferSizeInMB8/FlushIndexBufferSizeInMB
ColumnIndexSizeInKB64/ColumnIndexSizeInKB
MemtableThroughputInMB64/MemtableThroughputInMB
This sounds similar to /proc/sys/vm/swappiness misconfiguration. Is it zero
or close to zero? If setting it 0 solves your problem, make sure all your
nodes get this:
/etc/sysctl.conf:
vm.swappiness=0
On Wed, Apr 28, 2010 at 12:12 PM, Kyusik Chung kyu...@discovereads.comwrote:
Hello. I am
Isnt setting swappiness to a lower value a good idea only if you know you have
the physical RAM to support it? What Im observing on my box is that jsvc uses
up all the physical RAM. Its VM size is 4-5GB right now (not sure if it will
continue to grow).
Apologies if Im misunderstanding how