High CPU load

Hey all, I'm seeing really high cpu load across most of my cluster (4
m2.xlarge ec2 instances, 17gb ram). We also get timeouts on searching
despite search and index volume being in the magnitude of 1-5 per second.

Here's a dump from jstack on one of the servers and our
config: https://gist.github.com/ba4bc5609f71ede28132

I only have 5m documents at a size 35GB. Many requests just time out - we
turned on http logging with jetty and those requests just don't make it.

Any ideas for further debugging? Unfortunately I have no idea how to read a
java stack.

-- Courtenay

--

courtenay wrote:

Hey all, I'm seeing really high cpu load across most of my cluster
(4 m2.xlarge ec2 instances, 17gb ram). We also get timeouts on
searching despite search and index volume being in the magnitude of
1-5 per second.

Here's a dump from jstack on one of the servers and our
config: jstack · GitHub

I only have 5m documents at a size 35GB. Many requests just time
out - we turned on http logging with jetty and those requests just
don't make it.

Any ideas for further debugging? Unfortunately I have no idea how
to read a java stack.

The thread dump looks normal, although there could have been a few
filesystem operations hanging. Are your indices stored on EBS?

Are you able to isolate what queries are taking so long? Can you
post those or send them to me privately? Seems odd given your small
dataset.

-Drew

--