From 5ab609b3bb8aeeaceb28707df9ac955cce1c3c19 Mon Sep 17 00:00:00 2001 From: Jim Harris Date: Wed, 13 Feb 2019 13:16:51 -0700 Subject: [PATCH] test/qos: allow more variance due to iSCSI overhead Running iSCSI in loopback runs into some cases where a core is considerably impacted by the initiator TCP traffic, even though the scheduler is trying to schedule it on different cores. Gang is working through modifying the QoS tests to just use bdevperf instead of iSCSI, but until that is ready, let's allow more variance to reduce the fairly significant rate of test failures we're seeing due to this test. Signed-off-by: Jim Harris Change-Id: I99fe13621f5ca2f8e6402fc17fb859df815fe05e Reviewed-on: https://review.gerrithub.io/c/444458 Reviewed-by: Ben Walker Reviewed-by: Shuhei Matsumoto Reviewed-by: Changpeng Liu Tested-by: SPDK CI Jenkins --- test/iscsi_tgt/qos/qos.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/test/iscsi_tgt/qos/qos.sh b/test/iscsi_tgt/qos/qos.sh index fd4d94ff4..4a8d73ca8 100755 --- a/test/iscsi_tgt/qos/qos.sh +++ b/test/iscsi_tgt/qos/qos.sh @@ -32,7 +32,7 @@ function check_qos_works_well() { if [ $enable_limit = true ]; then #qos realization is related with bytes transfered.It currently have like 5% variation. - retval=$(echo "$qos_limit*0.95 < $read_result && $read_result < $qos_limit*1.05" | bc) + retval=$(echo "$qos_limit*0.85 < $read_result && $read_result < $qos_limit*1.05" | bc) if [ $retval -eq 0 ]; then echo "Failed to limit the io read rate of malloc bdev by qos" exit 1