New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Possible performance degradation using all CPUs in small instances #21

Open
eyalgutkind opened this Issue Feb 16, 2017 · 3 comments

Comments

Projects
None yet
3 participants
@eyalgutkind
Copy link

eyalgutkind commented Feb 16, 2017

Based on testing completed on C3.2xl instances. the CPU info is:
[centos@ip-172-31-27-168 ~]$ cat /etc/scylla.d/cpuset.conf
CPUSET="--cpuset 0-7 "

It possible that vCPU 0 usage cause performance degradation due to IRQ balancing.
/cc @glommer

@glommer

This comment has been minimized.

Copy link
Contributor

glommer commented Mar 5, 2017

This happens because the network script wrongly configures the instance in MQ instead of SQ mode. Vlad and myself are working on a fix for the main Scylla rpm, and the AMI should be updated as soon as this is fixed.

A temporary workaround is to configure the network manually in SQ mode and then adjust the cpuset manually to 1-7.

@tzach

This comment has been minimized.

Copy link

tzach commented Mar 5, 2017

A temporary workaround is to configure the network manually in SQ mode

How do you do that?

@glommer

This comment has been minimized.

Copy link
Contributor

glommer commented Mar 5, 2017

for the cpuset, just change it manually to 1-7 in cpuset.conf.
That will already bring most of the performance back.

As an extra step:

sudo sh /usr/lib/scylla/posix_net_conf.sh -sq 
sudo sed -i "s/SET_NIC=yes/SET_NIC=no/" /etc/sysconfig/scylla-server

The only problem is that posix_net_conf.sh -sq would have to be re-executed at every VM boot.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment