Add a slow e2e test to monitor kubelet resource usage

This test tracks kubelet resource usage over a long period of time (1hr)
when running N pods (e.g., N=0,50), and prints out the resource usage. This
would give us an idea how much kubelet's management overhead is in a stable
cluster.

Some followup items:
 * Use a more realistic workload (e.g., including probing)
 * Fail the test if the resource usage is too high.

Caveat:
 * We assume the scheduler would do a decent job distributing the pause pods,
   but we should double check.
 * Cluster addon pods could be unevenly distributed and skews the resource
   usage on nodes.
This commit is contained in:
Yu-Ju Hong
2015-09-22 16:38:36 -07:00
parent ab85b9a7d6
commit 0b8244505c
4 changed files with 131 additions and 18 deletions

View File

@@ -134,6 +134,7 @@ GCE_FLAKY_TESTS=(
GCE_SLOW_TESTS=(
"SchedulerPredicates\svalidates\sMaxPods\slimit " # 8 min, file: scheduler_predicates.go, PR: #13315
"Nodes\sResize" # 3 min 30 sec, file: resize_nodes.go, issue: #13323
"resource\susage\stracking" # 1 hour, file: kubelet_perf.go, slow by design
)
# Tests which are not able to be run in parallel.
@@ -147,6 +148,7 @@ GCE_PARALLEL_SKIP_TESTS=(
"SchedulerPredicates"
"Services.*restarting"
"Shell.*services"
"resource\susage\stracking"
)
# Tests which are known to be flaky when run in parallel.