Skip to content
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

os/bluestore/BlueFS: For logs of rocksdb & bluefs only use directio. #11012

Merged
merged 1 commit into from Sep 8, 2016

Conversation

majianpeng
Copy link
Member

Now it use bluefs_buffered_io to control whether use buffer or directio
when write. But in fact for logs of rocksdb & bluefs, whether
bluefs_buffer_io is true or false, the logs only need directio.

Signed-off-by: Jianpeng Ma jianpeng.ma@intel.com

Now it use bluefs_buffered_io to control whether use buffer or directio
when write. But in fact for logs of rocksdb & bluefs, whether
bluefs_buffer_io is true or false, the logs only need directio.

Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
@liewegas liewegas merged commit 73e4d2d into ceph:master Sep 8, 2016
@markhpc
Copy link
Member

markhpc commented Sep 9, 2016

This appears to be causing a ~10% performance regression with 4K random writes on NVMe devices when bluefs_buffer_io is set to true.

@markhpc
Copy link
Member

markhpc commented Sep 12, 2016

Upon further testing, it looks like the majority, if not all, of the effect is coming from the second part of the conditional, ie "h->writer_type == WRITER_WAL". I don't know why this would be the case however.

@liewegas
Copy link
Member

liewegas commented Sep 12, 2016 via email

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
3 participants