a) 0.6.1 is ancient, upgrade to 0.6.6 (see
http://www.riptano.com/blog/whats-new-cassandra-066 for links to all
the improvements since 0.6.1 -- the links to older versions are at the
bottom)

b) increase the memtable flush thresholds to reduce the need for
compaction (8x the defaults is a decent starting point, assuming you
are giving a reasonable amount of your 32GB to the JVM for heap)

c) set compaction priority lower
(http://www.riptano.com/blog/cassandra-annotated-changelog-063) to
make compaction hurt less

d) enable the dynamic snitch
(http://www.riptano.com/blog/whats-new-cassandra-065) to route around
nodes that are slow b/c of compaction

On Thu, Oct 14, 2010 at 3:23 PM, Henry Luo <h...@choicestream.com> wrote:
> We have a five node cluster, using replication factor of 3.  The application
> is only sending write requests at this point – we’d like to gain some
> operation experience with it first before start read from it.
>
>
>
> We are seeing over a hundred compaction activities on each server, some of
> them are for HintsColumnFamily.
>
>
>
> Each machine has 32 GB memory, two disk arrays, one with raid 0 for commit
> log, one with raid 5 for data. We are using version 0.6.1, and pretty much
> the out of the box storage.xml.
>
>
>
> Is this normal? Where should we look for tuning?
>
>
>
> Here is the ring info
>
>
>
> Address       Status     Load
> Range                                      Ring
>
>
> 103348149328693428942388257816272166328
>
> 10.100.10.68  Up         136.41 GB
> 62116456964768051843784433654721163092     |<--|
>
> 10.100.10.64  Up         136.31 GB
> 82105179051854269619799333828977372565     |   ^
>
> 10.100.10.66  Up         152.77 GB
> 92197953251627500070365755299174650936     v   |
>
> 10.100.10.72  Up         71.38 GB
> 102264937228017528105060257264614100661    |   ^
>
> 10.100.10.76  Up         24.8 GB
> 103348149328693428942388257816272166328    |-->|
>
>
>
> Thanks.
>
> Henry
>
>
>
> ________________________________
> The information transmitted is intended only for the person or entity to
> which it is addressed and may contain confidential, proprietary, and/or
> privileged material. Any review, retransmission, dissemination or other use
> of, or taking of any action in reliance upon this information by persons or
> entities other than the intended recipient is prohibited. If you received
> this in error, please contact the sender and delete the material from all
> computers.
>



-- 
Jonathan Ellis
Project Chair, Apache Cassandra
co-founder of Riptano, the source for professional Cassandra support
http://riptano.com

Reply via email to