cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Anuj Wadehra <>
Subject Re: Re:Re: Re: Re: Cassandra Tuning Issue
Date Wed, 09 Dec 2015 07:13:38 GMT
You just need to send mail to ..Everyone on mailing list including
you will get the mail..


Sent from Yahoo Mail on Android

From:"xutom" <>
Date:Wed, 9 Dec, 2015 at 7:28 am
Subject:Re:Re: Re: Re: Cassandra Tuning Issue

Hi Anuj,
Thanks! I will retry now!
By the way, how to " inform the C* email list as well so that others know" as Jack said? I
am sorry I have not do that yet. 


At 2015-12-09 01:09:07, "Anuj Wadehra" <> wrote:

Hi Jerry,

Its great that you got performance improvement. Moreover, I agree with what Graham said. I
think that you are using extremely large Heaps with CMS and that too in very odd ratio..Having
40G for new gen and leaving only 20G old gen seems unreasonable..Its hard to believe that
you are having reasonable Gc pauses..Please recheck..I would suggest you to test your performance
with much smaller heap..may be 16G max heap n 4G new gen..moreover make sure that you apply
all the recommended Production settings suggested by DataStax at

Dont worry about wasting your memory, it will be used for OS caching and you can get even
better performance..



Sent from Yahoo Mail on Android

From:"Jack Krupansky" <>
Date:Tue, 8 Dec, 2015 at 8:07 pm
Subject:Re: Re: Re: Cassandra Tuning Issue

Great! Make sure to inform the C* email list as well so that others know.

-- Jack Krupansky

On Tue, Dec 8, 2015 at 7:44 AM, xutom <> wrote:

Dear Jack,
    Thank you very much! Now we have much better performance when we insert the same partition
keys in the same batch.


At 2015-12-07 13:08:31, "Jack Krupansky" <> wrote:

If you combine inserts for multiple partition keys in the same batch you negate most of the
effect of token-aware routing. It's best to insert only rows with the same partition key in
a single batch. You also need to set the partition key for routing for the batch.

Also, RF=2 is not recommended since it does not permit quorum operations if a replica node
is down. RF=3 is generally more appropriate.

-- Jack Krupansky

On Sun, Dec 6, 2015 at 10:27 PM, xutom <> wrote:

Dear all,
    Thanks for ur reply!
    Now I`m using Apache Cassandra 2.1.1 and my JDK is 1.7.0_79,  my keyspace replication
factor is 2,and I do enable the "token aware". The GC configuration is default for such
# GC tuning options
JVM_OPTS="$JVM_OPTS -XX:+UseConcMarkSweepGC"
JVM_OPTS="$JVM_OPTS -XX:+CMSParallelRemarkEnabled"
    And I check the gc log: gc.log.0.current, I found there is only one Full GC. The stop-the-world
times is low.
CMS-initial-mark: 0.2747280 secs
CMS-remark: 0.3623090 secs

    The insert codes in my test client are following:
            String content = RandomStringUtils.randomAlphabetic(120);
            cluster = Cluster
                    .withCredentials("test", "test")
                    .withLoadBalancingPolicy(new TokenAwarePolicy(new DCAwareRoundRobinPolicy()))

            session = cluster.connect("demo");

            PreparedStatement insertPreparedStatement = session.prepare(
                        "   INSERT INTO teacher (id, lastname, firstname,
city) " +
                                "VALUES (?, ?, ?, ?); ");

            BatchStatement batch = new BatchStatement();
            for (; i < max; i+=5) {
                try {
                    batch.add(insertPreparedStatement.bind(i, "Entre Nous",
"adsfasdfa1", content));
                    batch.add(insertPreparedStatement.bind(i+1, "Entre Nous",
"adsfasdfa2", content));
                    batch.add(insertPreparedStatement.bind(i+2, "Entre Nous",
"adsfasdfa3", content));
                    batch.add(insertPreparedStatement.bind(i+3, "Entre Nous",
"adsfasdfa4", content));
                    batch.add(insertPreparedStatement.bind(i+4, "Entre Nous",
"adsfasdfa5", content));
//                    System.out.println("the is is " + i);

                    thisTimeCount += 5;

At 2015-12-07 00:40:06, "Graham Sanderson" <> wrote:

What version of C* are you using; what JVM version - you showed a partial GC config but if
that is still CMS (not G1) then you are going to have insane GC pauses... 

Depending on C* versions are you using on/off heap memtables and what type

Those are the sorts of issues related to fat nodes; I'd be worried about - we run very nicely
at 20G total heap and 8G new - the rest of our 128G memory is disk cache/mmap and all of the
off heap stuff so it doesn't go to waste

That said I think Jack is probably on the right path with overloaded coordinators- though
you'd still expect to see CPU usage unless your timeouts are too low for the load, In which
case the coordinator would be getting no responses in time and quite possibly the other nodes
are just dropping the mutations (since they don't get to them before they know the coordinator
would have timed out) - I forget the command to check dropped mutations off the top of my
head but you can see it in opcenter

If you have GC problems you certainly

Expect to see GC cpu usage but depending on how long you run your tests it might take you
a little while to run thru 40G

I'm personally not a fan off >32G (ish) heaps as you can't do compressed oops and also
it is unrealistic for CMS ... The word is that G1 is now working ok with C* especially on
newer C* and JDK versions, but that said it takes quite a lot of thru-put to require insane
quantities of young gen... We are guessing that when we remove all our legacy thrift batch
inserts we will need less - and as for 20G total we actually don't need that much (we dropped
from 24 when we moved memtables off heap, and believe we can drop further)

Sent from my iPhone

On Dec 6, 2015, at 9:07 AM, Jack Krupansky <> wrote:

What replication factor are you using? Even if your writes use CL.ONE, Cassandra will be attempting
writes to the replica nodes in the background.

Are your writes "token aware"? If not, the receiving node has the overhead of forwarding the
request to the node that owns the token for the primary key.

For the record, Cassandra is not designed and optimized for so-called "fat nodes". The design
focus is "commodity hardware" and "distributed cluster" (typically a dozen or more nodes.)

That said, it would be good if we had a rule of thumb for how many simultaneous requests a
node can handle, both external requests and inter-node traffic. I think there is an open Jira
to enforce a limit on inflight requests so that nodes don't overloaded and start failing in
the middle of writes as you seem to be seeing.

-- Jack Krupansky

On Sun, Dec 6, 2015 at 9:29 AM, jerry <> wrote:

Dear All,

    Now I have a 4 nodes Cassandra cluster, and I want to know the highest performance of
my Cassandra cluster. I write a JAVA client to batch insert datas into ALL 4 nodes Cassandra,
when I start less than 30 subthreads in my client applications to insert datas into cassandra,
it will be ok for everything, but when I start more than 80 or 100 subthreads in my client
applications, there will be too much timeout Exceptions (Such as: Cassandra timeout during
write query at consistency ONE (1 replica were required but only 0 acknowledged the write)).
And no matter how many subthreads or even I start multiple clients with multiple subthreads
on different computers, I can get the highest performance for about 60000 - 80000 TPS. By
the way, each row I insert into cassandra is about 130 Bytes.
    My 4 nodes of Cassandra is :
        CPU: 4*15
        Memory: 512G
        Disk: flash card (only one disk but better than SSD)
    My cassandra configurations are:
        MAX_HEAP_SIZE: 60G
        NEW_HEAP_SIZE: 40G

    When I insert datas into my cassandra cluster, each nodes has NOT reached bottleneck
such as CPU or Memory or Disk. Each of the three main hardwares is idle。So I think maybe
there is something wrong about my configuration of cassandra cluster. Can somebody please
help me to My Cassandra Tuning? Thanks in advances!




View raw message