Forums

Home / Forums

You need to log in to create posts and topics. Login · Register

Improve XFS performance on storage nodes


Mykola Perehinets
38 Posts
March 20, 2017, 11:22 am

Hi folks,
Please add on to default ceph.conf next line in section [osd]:
osd mkfs options xfs = -f -i size=4096 -d agcount=32
osd mount options xfs = rw,noatime,inode64,noquota,allocsize=4M,logbufs=8,logbsize=256k,largeio,swalloc
Detail info: http://docs.ceph.com/docs/jewel/rados/configuration/osd-config-ref/

And please add in crontab on all storage nodes next line:
......
0 5 * * * /usr/sbin/xfs_fsr -t 7200 > /dev/null 2>&1
......
This is can improve XFS performance on storage nodes.

Last edited on March 20, 2017, 11:23 am · #1

admin
2,921 Posts
March 20, 2017, 2:26 pm

Thank you very much. We will definitely look into this.

So far most  our config is close to the Ceph defaults, in the future we plan to allow the user to define/add their own configs and can select between different config templates that are optimized for specific usage.

Thanks !

#2

Mykola Perehinets
38 Posts
March 20, 2017, 3:36 pm

Ok. Thank you. But remember in future CEPH use BlueStore and this options does not needed in solution.

This option need for all users now!

Detail info:

https://www.sebastien-han.fr/blog/2016/03/21/ceph-a-new-store-is-coming/

Testing Ceph BlueStore with the Kraken release

http://events.linuxfoundation.org/sites/events/files/slides/LinuxCon%20NA%20BlueStore.pdf

http://ceph.com/planet/testing-ceph-bluestore-with-the-kraken-release/

Thank you again.

Last edited on March 21, 2017, 10:23 pm · #3