I installed solr-3.6 in my local windows box and it worked fine.

I installed solr-4.0 in amazon ec2 linux large instance and the cpu usage shot upto 100%. It maintained at 80-90% average cpu power.

I thought it could be because of 4.0, So I installed 3.6 in EC2 again. But again the CPU usage was 80-90% average.

With both the versions, solr works in EC2. dont know why CPU usage is so high. i started the solr server using "sudo nohup java -jar start.jar &"

In my local box java 1.7 is installed and in EC2 it is 1.6.0_24. I have mapped solr dir to an EBS volume.

/dev/mapper/vg1-solr   8361916   1935928   6342128  24% /home/ec2-user/SOLR/solr/example/solr

Is there any known issue ?

  • 1
    Well, how does your local box compare to the EC2 instance performance-wise, and what's the average CPU used by Solr on your local box? – ceejayoz Jul 3 '12 at 21:25

Looks like there is an issue, which has been outlined in the below response. For more details, have a look at Anyone else experiencing high rates of Linux server crashes during a leap second day?

We faced just this issue yesterday - the problem is because of the leap second on June 30 2012. A linux kernel component that manages sleep times isn't updated to the correct time and this causes extremely high CPU usage for Java processes. Related question on serverfault and my fix I derived from it (for Debian):

(issue these commands from the command line)

export LANG="en_EN"
date -s "`date`"

/etc/init.d/ntp stop
ntpdate pool.ntp.org
/etc/init.d/ntp start

For Red Hat derived systems, I believe you replace ntp with ntpd.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.