Alan Orth
3a5b50f941
All servers with non-rotating disks (SSDs) should be running noop, and the rest should be running deadline. Signed-off-by: Alan Orth <alan.orth@gmail.com>
21 lines
1.3 KiB
Django/Jinja
21 lines
1.3 KiB
Django/Jinja
# set noop scheduler for non-rotating disks (SSD)
|
|
# the glob matches sd* on physical hosts, xvd* on Xen hosts, and vd* on KVM hosts
|
|
ACTION=="add|change", KERNEL=="xvd[a-z]|[sv]d[a-z]", ATTR{queue/rotational}=="0", ATTR{queue/scheduler}="noop"
|
|
|
|
# set deadline scheduler for rotating disks
|
|
# the glob matches sd* on physical hosts, xvd* on Xen hosts, and vd* on KVM hosts
|
|
#
|
|
# newer KVM guests (Ubuntu 14.04 at least, so Linux >=3.13) using virtio vda don't
|
|
# allow it to be set, but it's non fatal so let's just set it for now
|
|
ACTION=="add|change", KERNEL=="xvd[a-z]|[sv]d[a-z]", ATTR{queue/rotational}=="1", ATTR{queue/scheduler}="deadline"
|
|
|
|
{% if 'storage' in group_names %}
|
|
## set disk tweaks
|
|
## see: http://www.fhgfs.com/wiki/wikka.php?wakka=StorageServerTuning
|
|
## see: http://community.gluster.org/a/linux-kernel-tuning-for-glusterfs/
|
|
# give the IO scheduler more flexibility by increasing the number of schedulable requests (queue_depth * 2):
|
|
ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/nr_requests}="512"
|
|
# To improve throughput for sequential reads, increase the maximum amount of read-ahead data. The actual amount of read-ahead is adaptive, so using a high value here won't harm performance for small random access.
|
|
ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/read_ahead_kb}="4096"
|
|
{% endif %}
|