ivampiresp/Spdk
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

test/scheduler: wait for scheduler after thread creation in busy()

Browse Source 
All other tests wait for scheduling to occur before verifying
the results. Even for 100% busy threads, they are first
round robined between cores.

In some edge cases, if threads are created right at the
end of the scheduling period - they will be 100% busy.
Yet due to low busy_tsc it will be deemed that two of such threads
can be placed on single core. To prevent that a few
full scheduling periods have to be observed.

Thus added sleep for 10 scheduling periods before verifying
the results, similar to other test cases in this file.

While here increased the sleep time between samples to
match the scheduling period. With 1 sec default scheduling period,
no change would occur between two 0.5 sec samples.

Fixes #2140

Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I5abf3bc04c6f65f4faae2bcd3e6828be72f97311
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9387
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
This commit is contained in:
Tomasz Zawadzki 2021-09-03 10:12:17 +02:00
parent 0c106d2b84
commit 84fa73eb5c
1 changed files with 4 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
test/scheduler/load_balancing.sh
View File

@ -16,6 +16,7 @@ busy() {
local selected_cpus cpu local selected_cpus cpu
local reactor_framework local reactor_framework
local threads thread local threads thread
local sched_period=1 # default, 1s
# Create two busy threads with two cpus (not including main cpu) and check if either of # Create two busy threads with two cpus (not including main cpu) and check if either of
# them is moved to either of the selected cpus. Expected load is ~100% on each thread and # them is moved to either of the selected cpus. Expected load is ~100% on each thread and
@ -26,11 +27,13 @@ busy() {
thread0=$(create_thread -n "thread0" -m "$(mask_cpus "${selected_cpus[@]}")" -a 100) thread0=$(create_thread -n "thread0" -m "$(mask_cpus "${selected_cpus[@]}")" -a 100)
thread1=$(create_thread -n "thread1" -m "$(mask_cpus "${selected_cpus[@]}")" -a 100) thread1=$(create_thread -n "thread1" -m "$(mask_cpus "${selected_cpus[@]}")" -a 100)
sleep $((10 * sched_period))
local samples=0 local samples=0
xtrace_disable xtrace_disable
while ((samples++ < 5)); do while ((samples++ < 5)); do
sleep 0.5s sleep $sched_period
all_set=0 all_set=0
reactor_framework=$(rpc_cmd framework_get_reactors | jq -r '.reactors[]') reactor_framework=$(rpc_cmd framework_get_reactors | jq -r '.reactors[]')