This is an automated email from the ASF dual-hosted git repository.
pbacsko pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/yunikorn-core.git
The following commit(s) were added to refs/heads/master by this push:
new 23214ea5 [YUNIKORN-2371] Add failed headroom checks to the allocation
log (#781)
23214ea5 is described below
commit 23214ea54db56b90edc50b78857c19a6c9152952
Author: Peter Bacsko <[email protected]>
AuthorDate: Thu Feb 1 00:54:25 2024 +0100
[YUNIKORN-2371] Add failed headroom checks to the allocation log (#781)
Closes: #781
Signed-off-by: Peter Bacsko <[email protected]>
---
pkg/scheduler/objects/application.go | 5 +++
pkg/scheduler/objects/application_test.go | 67 +++++++++++++++++++++++++++----
2 files changed, 65 insertions(+), 7 deletions(-)
diff --git a/pkg/scheduler/objects/application.go
b/pkg/scheduler/objects/application.go
index 8f1f49a5..17fb264a 100644
--- a/pkg/scheduler/objects/application.go
+++ b/pkg/scheduler/objects/application.go
@@ -57,6 +57,9 @@ var rateLimitedAppLog *log.RateLimitedLogger
const (
Soft string = "Soft"
Hard string = "Hard"
+
+ NotEnoughUserQuota = "Not enough user quota"
+ NotEnoughQueueQuota = "Not enough queue quota"
)
type PlaceholderData struct {
@@ -951,6 +954,7 @@ func (sa *Application) tryAllocate(headRoom
*resources.Resource, allowPreemption
// NOTE: preemption most likely will not help in this case. The
chance that preemption helps is mall
// as the preempted allocation must be for the same user in a
different queue in the hierarchy...
if !userHeadroom.FitInMaxUndef(request.GetAllocatedResource()) {
+ request.LogAllocationFailure(NotEnoughUserQuota, true)
// error message MUST be constant!
continue
}
@@ -970,6 +974,7 @@ func (sa *Application) tryAllocate(headRoom
*resources.Resource, allowPreemption
}
}
sa.appEvents.sendAppDoesNotFitEvent(request, headRoom)
+ request.LogAllocationFailure(NotEnoughQueueQuota, true)
// error message MUST be constant!
continue
}
diff --git a/pkg/scheduler/objects/application_test.go
b/pkg/scheduler/objects/application_test.go
index 684f50c4..b8286ff2 100644
--- a/pkg/scheduler/objects/application_test.go
+++ b/pkg/scheduler/objects/application_test.go
@@ -39,6 +39,15 @@ import (
"github.com/apache/yunikorn-scheduler-interface/lib/go/si"
)
+var (
+ nilNodeIterator = func() NodeIterator {
+ return nil
+ }
+ nilGetNode = func(string) *Node {
+ return nil
+ }
+)
+
func setupUGM() {
userManager := ugm.GetUserManager()
userManager.ClearUserTrackers()
@@ -2272,13 +2281,7 @@ func TestAppDoesNotFitEvent(t *testing.T) {
app.sortedRequests = sr
attempts := 0
- app.tryAllocate(headroom, true, time.Second, &attempts, func()
NodeIterator {
- return nil
- }, func() NodeIterator {
- return nil
- }, func(s string) *Node {
- return nil
- })
+ app.tryAllocate(headroom, true, time.Second, &attempts,
nilNodeIterator, nilNodeIterator, nilGetNode)
noEvents := 0
err = common.WaitFor(10*time.Millisecond, time.Second, func() bool {
@@ -2294,6 +2297,56 @@ func TestAppDoesNotFitEvent(t *testing.T) {
assert.Equal(t, "alloc-0", records[1].ObjectID)
}
+func TestAllocationFailures(t *testing.T) {
+ setupUGM()
+
+ res, err := resources.NewResourceFromConf(map[string]string{"memory":
"100", "vcores": "10"})
+ assert.NilError(t, err)
+ headroom, err :=
resources.NewResourceFromConf(map[string]string{"memory": "0", "vcores": "0"})
+ assert.NilError(t, err)
+ ask := newAllocationAsk("alloc-0", "app-1", res)
+ app := newApplication(appID1, "default", "root")
+ queue, err := createRootQueue(nil)
+ assert.NilError(t, err, "queue create failed")
+ app.queue = queue
+ sr := sortedRequests{}
+ sr.insert(ask)
+ app.sortedRequests = sr
+ attempts := 0
+
+ // case #1: not enough queue headroom
+ app.tryAllocate(headroom, true, time.Second, &attempts,
nilNodeIterator, nilNodeIterator, nilGetNode)
+ assert.Equal(t, 1, len(ask.allocLog))
+ assert.Equal(t, int32(1), ask.allocLog[NotEnoughQueueQuota].Count)
+
+ // case #2: not enough user quota
+ // create config with resource limits for "testuser"
+ conf := configs.QueueConfig{
+ Name: "root",
+ Parent: true,
+ SubmitACL: "*",
+ Limits: []configs.Limit{
+ {
+ Limit: "leaf queue limit",
+ Users: []string{
+ "testuser",
+ },
+ MaxResources: map[string]string{
+ "memory": "1",
+ "vcores": "1",
+ },
+ },
+ },
+ }
+ err = ugm.GetUserManager().UpdateConfig(conf, "root")
+ assert.NilError(t, err)
+ headroom, err =
resources.NewResourceFromConf(map[string]string{"memory": "1000", "vcores":
"1000"})
+ assert.NilError(t, err)
+ app.tryAllocate(headroom, true, time.Second, &attempts,
nilNodeIterator, nilNodeIterator, nilGetNode)
+ assert.Equal(t, 2, len(ask.allocLog))
+ assert.Equal(t, int32(1), ask.allocLog[NotEnoughUserQuota].Count)
+}
+
func TestGetOutstandingRequests(t *testing.T) {
// Create a sample Resource and AllocationAsk
resMap := map[string]string{"memory": "100", "vcores": "10"}
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
