I gave MySQL 5.1.38 a try (that's not the clustered version, so nevermind what
happens if a disk is lost) and I saw the same persistently increasing latency
that I saw with Cassandra.
I also tried storing into files on the local filesystem (not clustered or
transactional, so nevermind what
and preferably trunk if you are
stress testing
-Jonathan
On Tue, Dec 1, 2009 at 12:11 PM, Freeman, Tim tim.free...@hp.com wrote:
In an 8 hour test run, I've seen the read latency for Cassandra drift fairly
linearly from ~460ms to ~900ms. Eventually my application gets starved for
reads
again?
On Thu, Dec 3, 2009 at 12:58 PM, Freeman, Tim tim.free...@hp.com wrote:
I ran another test last night with the build dated 29 Nov 2009. Other than
the Cassandra version, the setup was the same as before. I got qualitatively
similar results as before, too -- the read latency increased
I suppose the point behind RackAware is that you can lose one rack or
datacenter and then recover somehow. How would the recovery work? I'd
probably have part of my cluster left, the replication factor would be three,
and I'd have less than a quorum for some of the rows. I'd have to add a
I use
RpcTimeoutInMillis3/RpcTimeoutInMillis
in storage-conf.xml, and for the most part that makes the timeouts go away. My
application would rather wait 30 seconds if needed, since it would otherwise
retry. YMMV. I hear that Cassandra 0.5 will be better at avoiding the
timeouts.
I'm not going to be on Amazon, but I'm planning to use hostnames instead of
IP's and a dynamically generated /etc/hosts file and I think that would deal
with this problem. I'm sure a private DNS server would be just as good.
My real motive in saying this is so someone will scream at me if I'm
I'd like to be able to say that a cluster of Cassandra servers keeps working
even if one of the nodes goes down. This would require the clients to connect
to a different node if the one that it's using goes down. Is there code on the
client side to do this already?
There's a Cassandra.Client
I'm running 0.4.1. I used to get timeouts, then I changed my timeout from 5
seconds to 30 seconds and I get no more timeouts. The relevant line from
storage-conf.xml is:
RpcTimeoutInMillis3/RpcTimeoutInMillis
The maximum latency is often just over 5 seconds in the worst case when I
.
But If I've multiple servers and have inserted the data with ONE then what it's
means I need to use only QUORUM instead ?
Thanks
On Mon, Nov 16, 2009 at 7:27 PM, Freeman, Tim
tim.free...@hp.commailto:tim.free...@hp.com wrote:
You might want to try ConsistencyLevel.QUORUM instead
The symptoms are consistent with
https://issues.apache.org/jira/browse/CASSANDRA-533. I agree with Jonathan
that the fix for that is in Trunk and not in 0.4.2. There's a code fix given
in the bug, if you want to hack source and recompile.
Tim Freeman
Email: tim.free...@hp.com
Desk in Palo
10 matches
Mail list logo