Update io-scheduler rules for rootdisk based on performance analysis.
Previous settings resulted in sm etcd failures when system was under I/O
stress. Tune to improve latency of critical disk I/O (etcdserver), and
responsiveness of kube-apiserver.
Changes are to let slice_idle go back to default of 8 and to enable
strict_guarantees.
Test Plan
PASS: Install system and confirm scheduler options
PASS: I/O stress soak - verify no stability or performance issues
PASS: Drbd-invalidate - verify successful resync
Closes-Bug: 1953074
Change-Id: I2c42b8850e844f24f53013386c9838eac260bf30
Signed-off-by: Gerry Kopec <gerry.kopec@windriver.com>
94b9530cd9