With a single node I get 3K for cassandra 1.0.12 and 1.2.12. So I suspect
there is some network chatter. I have started looking at the sources, hoping
to find something.
1.2 is pretty stable, I doubt there is anything in there that makes it run
slower than 1.0. It’s probably something in
On Wed, Dec 11, 2013 at 10:49 PM, Aaron Morton aa...@thelastpickle.comwrote:
It is the write latency, read latency is ok. Interestingly the latency is
low when there is one node. When I join other nodes the latency drops about
1/3. To be specific, when I start sending traffic to the other
Why did you switch to RandomPartitioner away from Murmur3Partitioner? Have
you tried with Murmur3?
1. # partitioner: org.apache.cassandra.dht.Murmur3Partitioner
2. partitioner: org.apache.cassandra.dht.RandomPartitioner
On Fri, Dec 6, 2013 at 10:36 AM, srmore comom...@gmail.com wrote:
On Thu, Dec 12, 2013 at 11:15 AM, J. Ryan Earl o...@jryanearl.us wrote:
Why did you switch to RandomPartitioner away from Murmur3Partitioner?
Have you tried with Murmur3?
1. # partitioner: org.apache.cassandra.dht.Murmur3Partitioner
2. partitioner:
Quote from
http://www.datastax.com/dev/blog/performance-improvements-in-cassandra-1-2
*Murmur3Partitioner is NOT compatible with RandomPartitioner, so if you’re
upgrading and using the new cassandra.yaml file, be sure to change the
partitioner back to RandomPartitioner*
On Thu, Dec 12, 2013 at
Changed memtable_total_space_in_mb to 1024 still no luck.
Reducing memtable_total_space_in_mb will increase the frequency of flushing to
disk, which will create more for compaction to do and result in increased IO.
You should return it to the default.
when I send traffic to one node its
Thanks Aaron
On Wed, Dec 11, 2013 at 8:15 PM, Aaron Morton aa...@thelastpickle.comwrote:
Changed memtable_total_space_in_mb to 1024 still no luck.
Reducing memtable_total_space_in_mb will increase the frequency of
flushing to disk, which will create more for compaction to do and result in
It is the write latency, read latency is ok. Interestingly the latency is low
when there is one node. When I join other nodes the latency drops about 1/3.
To be specific, when I start sending traffic to the other nodes the latency
for all the nodes increases, if I stop traffic to other
We have a 3 node cluster running cassandra 1.2.12, they are pretty big
machines 64G ram with 16 cores, cassandra heap is 8G.
The interesting observation is that, when I send traffic to one node its
performance is 2x more than when I send traffic to all the nodes. We ran
1.0.11 on the same box and
Hard to say much without knowing about the cassandra configurations.
Yes compactions/GC's could skipe the CPU, I had similar behavior with my
setup.
-VK
On Fri, Dec 6, 2013 at 7:40 PM, srmore comom...@gmail.com wrote:
We have a 3 node cluster running cassandra 1.2.12, they are pretty big
On Fri, Dec 6, 2013 at 9:32 AM, Vicky Kak vicky@gmail.com wrote:
Hard to say much without knowing about the cassandra configurations.
The cassandra configuration is
-Xms8G
-Xmx8G
-Xmn800m
-XX:+UseParNewGC
-XX:+UseConcMarkSweepGC
-XX:+CMSParallelRemarkEnabled
-XX:SurvivorRatio=4
Hi srmore,
Perhaps if you use jconsole and connect to the jvm using jmx. Then uner
MBeans tab, start inspecting the GC metrics.
/Jason
On Fri, Dec 6, 2013 at 11:40 PM, srmore comom...@gmail.com wrote:
On Fri, Dec 6, 2013 at 9:32 AM, Vicky Kak vicky@gmail.com wrote:
Hard to say much
You have passed the JVM configurations and not the cassandra configurations
which is in cassandra.yaml.
The spikes are not that significant in our case and we are running the
cluster with 1.7 gb heap.
Are these spikes causing any issue at your end?
On Fri, Dec 6, 2013 at 9:10 PM, srmore
On Fri, Dec 6, 2013 at 9:59 AM, Vicky Kak vicky@gmail.com wrote:
You have passed the JVM configurations and not the cassandra
configurations which is in cassandra.yaml.
Apologies, was tuning JVM and that's what was in my mind.
Here are the cassandra settings http://pastebin.com/uN42GgYT
Can you set the memtable_total_space_in_mb value, it is defaulting to 1/3
which is 8/3 ~ 2.6 gb in capacity
http://www.datastax.com/dev/blog/whats-new-in-cassandra-1-0-improved-memory-and-disk-space-management
The flushing of 2.6 gb to the disk might slow the performance if frequently
called, may
Looks like I am spending some time in GC.
java.lang:type=GarbageCollector,name=ConcurrentMarkSweep
CollectionTime = 51707;
CollectionCount = 103;
java.lang:type=GarbageCollector,name=ParNew
CollectionTime = 466835;
CollectionCount = 21315;
On Fri, Dec 6, 2013 at 9:58 AM, Jason Wee
Since how long the server had been up, hours,days,months?
On Fri, Dec 6, 2013 at 10:41 PM, srmore comom...@gmail.com wrote:
Looks like I am spending some time in GC.
java.lang:type=GarbageCollector,name=ConcurrentMarkSweep
CollectionTime = 51707;
CollectionCount = 103;
Not long: Uptime (seconds) : 6828
Token: 56713727820156410577229101238628035242
ID : c796609a-a050-48df-bf56-bb09091376d9
Gossip active: true
Thrift active: true
Native Transport active: false
Load : 49.71 GB
Generation No: 1386344053
Uptime
Changed memtable_total_space_in_mb to 1024 still no luck.
On Fri, Dec 6, 2013 at 11:05 AM, Vicky Kak vicky@gmail.com wrote:
Can you set the memtable_total_space_in_mb value, it is defaulting to 1/3
which is 8/3 ~ 2.6 gb in capacity
19 matches
Mail list logo