Skip to content

xdsclient: create LRSClient at time of initialisation #8483

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 12 commits into from
Aug 26, 2025
Merged

xdsclient: create LRSClient at time of initialisation #8483

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
e035e08
create lrs at creation time
eshitachandwani Jul 30, 2025
339ac57
test race
eshitachandwani Aug 1, 2025
8f53b14
add waitgroup
eshitachandwani Aug 1, 2025
56d3a58
error
eshitachandwani Aug 1, 2025
2f6d3b4
remove nil checks
eshitachandwani Aug 1, 2025
90bc53c
format
eshitachandwani Aug 5, 2025
40674fc
optional node id
eshitachandwani Aug 13, 2025
51309ec
optional node id
eshitachandwani Aug 13, 2025
c0ebd99
e2e test
eshitachandwani Aug 20, 2025
b8905c8
comments
eshitachandwani Aug 21, 2025
72427cb
merge
eshitachandwani Aug 21, 2025
c8c3ee4
final comments
eshitachandwani Aug 26, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion xds/internal/xdsclient/clientimpl.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -129,7 +129,14 @@ func newClientImpl(config *bootstrap.Config, metricsRecorder estats.MetricsRecor
if err != nil {
return nil, err
}
c := &clientImpl{XDSClient: client, xdsClientConfig: gConfig, bootstrapConfig: config, target: target, refCount: 1}
lrsC, err := lrsclient.New(lrsclient.Config{
Node: gConfig.Node,
TransportBuilder: gConfig.TransportBuilder,
})
if err != nil {
return nil, err
Copy link
Contributor

@purnesh42H purnesh42H Aug 1, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Now that we have moved it out of report load, we might have think a bit more here. Should error in lrs client creation be fatal? because not everyone is going to use internal xdsclient for load reporting.

Copy link
Contributor

@arjan-bal arjan-bal Aug 4, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In my opinion, we should keep the same behaviour as before the xDS client migration changes. If there are certain users who need to ignore LRS client creation failures, we can create a new issue to discuss if the bahviour changes makes sense.

Copy link
Contributor

@easwars easwars Aug 4, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it makes sense to retain the behavior of making LRS client creation failures be fatal. But we might want to change one minor thing in lrsclient.New. Currently it fails if node ID is empty in the configuration. We recently removed that check for the xDS client creation. I'm guessing other languages might not treat this as fatal for LRS creation.

@eshitachandwani : Could you please check what the other languages do and if required remove the check for empty node ID in lrsclient.New. Thanks.

Copy link
Member Author

@eshitachandwani eshitachandwani Aug 5, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If I understand correctly , Java is checking for not null here

Copy link
Contributor

@easwars easwars Aug 6, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That check is for the whole node proto or struct. Not just the node ID field.

Copy link
Member Author

@eshitachandwani eshitachandwani Aug 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry , that was a mistake , I think this is where they check for node id not null in Java
https://github.com/grpc/grpc-java/blob/f50726d32e216746642513add28e086094ce5506/xds/src/main/java/io/grpc/xds/client/EnvoyProtoData.java#L79

Copy link
Contributor

@arjan-bal arjan-bal Aug 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That is a private constructor. It's being called from the builder below which sets the id to an empty string by default. In Java empty and null strings are different.

Copy link
Member Author

@eshitachandwani eshitachandwani Aug 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ahh! Okay! Got it!

}
c := &clientImpl{XDSClient: client, xdsClientConfig: gConfig, bootstrapConfig: config, target: target, refCount: 1, lrsClient: lrsC}
c.logger = prefixLogger(c)
return c, nil
}
Expand Down
12 changes: 0 additions & 12 deletions xds/internal/xdsclient/clientimpl_loadreport.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,18 +32,6 @@ import (
//
// It returns a lrsclient.LoadStore for the user to report loads.
func (c *clientImpl) ReportLoad(server *bootstrap.ServerConfig) (*lrsclient.LoadStore, func(context.Context)) {
if c.lrsClient == nil {
lrsC, err := lrsclient.New(lrsclient.Config{
Node: c.xdsClientConfig.Node,
TransportBuilder: c.xdsClientConfig.TransportBuilder,
})
if err != nil {
c.logger.Warningf("Failed to create an lrs client to the management server to report load: %v", server, err)
return nil, func(context.Context) {}
}
c.lrsClient = lrsC
}

load, err := c.lrsClient.ReportLoad(clients.ServerIdentifier{
ServerURI: server.ServerURI(),
Extensions: grpctransport.ServerIdentifierExtension{
Expand Down
34 changes: 34 additions & 0 deletions xds/internal/xdsclient/tests/loadreport_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@ import (
"encoding/json"
"fmt"
"net"
"sync"
"testing"

"github.com/google/go-cmp/cmp"
Expand Down Expand Up @@ -437,3 +438,36 @@ func (s) TestReportLoad_StreamCreation(t *testing.T) {
defer sCancel3()
cancel3(sCtx3)
}

// TestConcurrentReportLoad verifies that the client can safely handle concurrent
// requests to initiate load reporting streams. It launches multiple goroutines
// that all call client.ReportLoad simultaneously.
func (s) TestConcurrentReportLoad(t *testing.T) {
ctx, cancel := context.WithTimeout(context.Background(), defaultTestTimeout)
defer cancel()

// Create a management server that serves LRS.
mgmtServer := e2e.StartManagementServer(t, e2e.ManagementServerOptions{SupportLoadReportingService: true})

// Create an xDS client with bootstrap pointing to the above server.
nodeID := uuid.New().String()
bc := e2e.DefaultBootstrapContents(t, nodeID, mgmtServer.Address)
client := createXDSClient(t, bc)

serverConfig, err := bootstrap.ServerConfigForTesting(bootstrap.ServerConfigTestingOptions{URI: mgmtServer.Address})
if err != nil {
t.Fatalf("Failed to create server config for testing: %v", err)
}
// Call ReportLoad() concurrently from multiple go routines.
var wg sync.WaitGroup
const numGoroutines = 10
wg.Add(numGoroutines)
for i := 0; i < numGoroutines; i++ {
go func() {
defer wg.Done()
_, cancelStore := client.ReportLoad(serverConfig)
Copy link
Contributor

@easwars easwars Aug 4, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would it make sense to have a loop here as well?

Copy link
Member Author

@eshitachandwani eshitachandwani Aug 5, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I am not sure if I understand, loop for what?

Copy link
Contributor

@easwars easwars Aug 6, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A loop inside the goroutine to start reporting load and subsequently canceling it. What I'm asking for is:

	for i := 0; i < numGoroutines; i++ {
		go func() {
			defer wg.Done()
			for j := 0; j < 100; j++ {
				_, cancelStore := client.ReportLoad(serverConfig)
				cancelStore(ctx)
			}
		}()
	}

Copy link
Member Author

@eshitachandwani eshitachandwani Aug 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You mean several ReportLoad() calls from one goroutine? Is this is better repro the real life case or to increase the chances of catching the race?

defer cancelStore(ctx)
}()
}
wg.Wait()
}