K8SPG-901: allow running without pgbouncer

[mayankshah1607]

CHANGE DESCRIPTION

Problem:
Operator does not allow running a cluster without PG bouncer. Specifying nil for .spec.proxy or .spec.proxy.pgbouncer still results in the default pgbouncer settings being applied.

Cause:
Operator adds default pgbouncer settings when set to nil.

Solution:
Honor nil setting for .spec.proxy, and apply pgbouncer defaults only when .spec.proxy is not nil

CHECKLIST

Jira

• Is the Jira ticket created and referenced properly?
• Does the Jira ticket have the proper statuses for documentation (Needs Doc) and QA (Needs QA)?
• Does the Jira ticket link to the proper milestone (Fix Version field)?

Tests

• Is an E2E test/test case added for the new feature/change?
• Are unit tests added where appropriate?

Config/Logging/Testability

• Are all needed new/changed options added to default YAML files?
• Are all needed new/changed options added to the Helm Chart?
• Did we add proper logging messages for operator actions?
• Did we ensure compatibility with the previous version or cluster upgrade process?
• Does the change support oldest and newest supported PG version?
• Does the change support oldest and newest supported Kubernetes version?

[Copilot]

Pull request overview

This PR enables running the Percona PostgreSQL Operator without PGBouncer by making the Proxy specification optional. Previously, the operator always initialized Proxy and PGBouncer objects even when not configured. Now, clusters can run with direct connections to PostgreSQL pods instead of routing through PGBouncer.

Changes:

• Modified the Default() method to only initialize PGBouncer when Proxy is configured
• Updated getHost() to return the pod service FQDN when Proxy/PGBouncer is not configured
• Added nil check for Proxy before accessing PGBouncer.EnvFrom (partially - bug found)

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
pkg/apis/pgv2.percona.com/v2/perconapgcluster_types.go	Modified Default() to conditionally initialize PGBouncer only when Proxy is non-nil
percona/controller/pgcluster/status.go	Updated getHost() to fallback to ClusterPodService when Proxy/PGBouncer is nil, improving service name handling
percona/controller/pgcluster/controller.go	Added partial nil check for Proxy before accessing PGBouncer.EnvFrom (missing PGBouncer check)

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 3 comments.

[Copilot]

1. Problem: When proxy/pgBouncer is not configured, getHost always returns the in-cluster primary Service FQDN, ignoring whether the primary Service is exposed as a LoadBalancer.
2. Why it matters: .status.host is printed as the user-facing “Endpoint”; without pgBouncer, users exposing Postgres via .spec.expose.type=LoadBalancer will still see an internal-only DNS name.
3. Fix: Mirror the existing LoadBalancer logic for the primary Service when proxy is nil: if .spec.expose is LoadBalancer, fetch the primary Service and prefer its ingress hostname/IP; otherwise return the Service FQDN.

Suggested change

	return svcFQDN(naming.ClusterPrimaryService(postgresCluster).Name, postgresCluster.Namespace), nil
	svcName := naming.ClusterPrimaryService(postgresCluster).Name
	// If the primary is exposed via LoadBalancer, prefer its external hostname/IP.
	if cr.Spec.Expose != nil && cr.Spec.Expose.Type == string(corev1.ServiceTypeLoadBalancer) {
	svc := &corev1.Service{}
	err := r.Client.Get(ctx, types.NamespacedName{Namespace: cr.Namespace, Name: svcName}, svc)
	if err != nil {
	return "", errors.Wrapf(err, "get %s service", svcName)
	}
	var host string
	for _, i := range svc.Status.LoadBalancer.Ingress {
	if len(i.Hostname) > 0 {
	host = i.Hostname
	} else if len(i.IP) > 0 {
	host = i.IP
	}
	}
	if host != "" {
	return host, nil
	}
	}
	return svcFQDN(svcName, postgresCluster.Namespace), nil

[Copilot]

1. Problem: The test suite covers Proxy: nil but does not cover the newly-supported case where Proxy is non-nil while Proxy.PGBouncer is nil (e.g. YAML proxy: {} or pgBouncer: null).
2. Why it matters: That is a realistic user configuration for disabling pgBouncer, and it would have caught the current ToCrunchy nil-dereference risk.
3. Fix: Add a ToCrunchy table entry that sets Spec.Proxy = &PGProxySpec{PGBouncer:nil} (or an empty PGProxySpec{}) and assert actual.Spec.Proxy == nil (and no panic/error).

[Copilot]

1. Problem: PGProxySpec.ToCrunchy will panic when .spec.proxy is present but .spec.proxy.pgBouncer is nil, because it later dereferences p.PGBouncer without a nil guard.
2. Why it matters: With this PR, Default() can leave Proxy non-nil while PGBouncer is nil (e.g. proxy: {}), making this crash reachable during PerconaPGCluster.ToCrunchy.
3. Fix: In ToCrunchy, treat p == nil || p.PGBouncer == nil as “proxy disabled” and return nil (or otherwise guard the dereference) before calling into p.PGBouncer.

Suggested change

	if p == nil {
	if p == nil || p.PGBouncer == nil {