OCPBUGS-82191: Surface async GCP instance creation errors - #160
OCPBUGS-82191: Surface async GCP instance creation errors#160RadekManak wants to merge 6 commits into
Conversation
openshift-ci-robot
added
the
jira/valid-reference
|
@RadekManak: This pull request references Jira Issue OCPBUGS-82191, which is invalid:
Comment The bug has been updated to refer to the pull request using the external bug tracker. DetailsIn response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the openshift-eng/jira-lifecycle-plugin repository. |
openshift-ci-robot
added
the
jira/invalid-bug
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
WalkthroughThe machine reconciler now checks recent GCP insert operations before instance creation, requeues pending and conflict cases, records provider operation failures, and handles creation-time not-found responses as transient. Compute service interfaces, mocks, and tests support these flows. ChangesGCP asynchronous instance creation
Estimated code review effort: 4 (Complex) | ~45 minutes Suggested reviewers: Sequence Diagram(s)sequenceDiagram
participant Reconciler
participant ZoneOperationsList
participant InstancesInsert
participant InstanceState
Reconciler->>ZoneOperationsList: Find latest matching insert operation
ZoneOperationsList-->>Reconciler: Return operation status and errors
alt Pending operation
Reconciler->>Reconciler: Requeue
else No pending operation
Reconciler->>InstancesInsert: Insert instance
InstancesInsert-->>Reconciler: Return operation or 409 conflict
Reconciler->>InstanceState: Reconcile instance state
end
Important Pre-merge checks failedPlease resolve all errors before merging. Addressing warnings is optional. ❌ Failed checks (1 error, 1 warning, 1 inconclusive)
✅ Passed checks (12 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
|
/jira refresh |
openshift-ci-robot
added
jira/valid-bug
|
@RadekManak: This pull request references Jira Issue OCPBUGS-82191, which is valid. The bug has been moved to the POST state. 3 validation(s) were run on this bug
DetailsIn response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the openshift-eng/jira-lifecycle-plugin repository. |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
pkg/cloud/gcp/actuators/machine/reconciler.go (1)
462-470: Record failed-create metrics for async operation failures too.This new return path updates conditions, but it bypasses
metrics.RegisterFailedInstanceCreate, so async provider-side failures won't show up in the existing failed-create metrics/alerts.📈 Proposed fix
if err := operationError(insertOp); err != nil { + metrics.RegisterFailedInstanceCreate(&metrics.MachineLabels{ + Name: r.machine.Name, + Namespace: r.machine.Namespace, + Reason: "failed to create instance via compute service", + }) r.providerStatus.Conditions = reconcileConditions(r.providerStatus.Conditions, metav1.Condition{ Type: string(machinev1.MachineCreated), Reason: machineCreationFailedReason, Message: err.Error(), Status: metav1.ConditionFalse, })🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@pkg/cloud/gcp/actuators/machine/reconciler.go` around lines 462 - 470, The return path when operationError(insertOp) is true updates r.providerStatus.Conditions but skips recording failed-create metrics; call metrics.RegisterFailedInstanceCreate (with the same labels/context used elsewhere for machine failures) immediately before returning from the operationError(insertOp) branch so async/provider-side failures are counted, keeping the existing condition update; locate the operationError(insertOp) check in reconciler.go and add the metrics.RegisterFailedInstanceCreate invocation prior to the fmt.Errorf return.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@pkg/cloud/gcp/actuators/machine/reconciler.go`:
- Around line 423-430: The current logic in reconciler.go's
latestVisibleInsertOperation check only skips re-create when existingOp.Status
!= "DONE" but allows a DONE operation with provider-side errors to fall through
to InstancesInsert; update the check in the reconciler (function
latestVisibleInsertOperation usage) to detect when existingOp.Status == "DONE"
and existingOp.Error != nil && len(existingOp.Error.Errors) > 0 and treat that
as a terminal failure: log the error (including existingOp.Error.Errors and
r.machine.Name), surface or return an appropriate non-retry error instead of
proceeding to InstancesInsert, and ensure any Requeue/Failure state prevents
issuing a new InstancesInsert for that same visible insert operation.
---
Nitpick comments:
In `@pkg/cloud/gcp/actuators/machine/reconciler.go`:
- Around line 462-470: The return path when operationError(insertOp) is true
updates r.providerStatus.Conditions but skips recording failed-create metrics;
call metrics.RegisterFailedInstanceCreate (with the same labels/context used
elsewhere for machine failures) immediately before returning from the
operationError(insertOp) branch so async/provider-side failures are counted,
keeping the existing condition update; locate the operationError(insertOp) check
in reconciler.go and add the metrics.RegisterFailedInstanceCreate invocation
prior to the fmt.Errorf return.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository: openshift/coderabbit/.coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: d4c33b05-57d3-4057-b16a-74abdfd9d1eb
📒 Files selected for processing (4)
pkg/cloud/gcp/actuators/machine/reconciler.gopkg/cloud/gcp/actuators/machine/reconciler_test.gopkg/cloud/gcp/actuators/services/compute/computeservice.gopkg/cloud/gcp/actuators/services/compute/computeservice_mock.go
3441b64 to
8e39814
Compare
|
@RadekManak: This pull request references Jira Issue OCPBUGS-82191, which is valid. 3 validation(s) were run on this bug
DetailsIn response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the openshift-eng/jira-lifecycle-plugin repository. |
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@pkg/cloud/gcp/actuators/machine/reconciler.go`:
- Around line 504-506: The current check using isNotFoundError(...) only matches
a bare *googleapi.Error and misses wrapped 404 errors; update the logic so the
code uses errors.As to detect a *googleapi.Error (or modify isNotFoundError to
internally use errors.As) and then checks err.Code == 404 before returning
&machinecontroller.RequeueAfterError{RequeueAfter: requeueAfterSeconds *
time.Second}; apply this change for all spots that currently call
isNotFoundError for requeue handling (the 404 requeue branches in reconciler.go,
e.g., the places around the RequeueAfterError returns) so wrapped errors are
handled consistently like the 409 handling that already uses errors.As.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository: openshift/coderabbit/.coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 0f81abaf-5607-41b7-b590-375f1cedc8de
📒 Files selected for processing (4)
pkg/cloud/gcp/actuators/machine/reconciler.gopkg/cloud/gcp/actuators/machine/reconciler_test.gopkg/cloud/gcp/actuators/services/compute/computeservice.gopkg/cloud/gcp/actuators/services/compute/computeservice_mock.go
🚧 Files skipped from review as they are similar to previous changes (3)
- pkg/cloud/gcp/actuators/services/compute/computeservice.go
- pkg/cloud/gcp/actuators/services/compute/computeservice_mock.go
- pkg/cloud/gcp/actuators/machine/reconciler_test.go
8e39814 to
f5b6f13
Compare
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@pkg/cloud/gcp/actuators/machine/reconciler.go`:
- Around line 504-506: reconcileMachineWithCloudState currently treats an
InstancesGet 404 as a transient condition and always returns a
RequeueAfterError, which causes endless retries when invoked from update();
change reconcileMachineWithCloudState signature to accept a boolean (e.g.,
treatNotFoundAsTransient bool) and only return
&machinecontroller.RequeueAfterError{...} when isNotFoundError(err) &&
treatNotFoundAsTransient is true; update all call sites so create() invokes
reconcileMachineWithCloudState(..., true) and update() invokes
reconcileMachineWithCloudState(..., false), leaving isNotFoundError and
RequeueAfterError logic intact otherwise.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository: openshift/coderabbit/.coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: d9048dd5-fbe5-4632-917f-04e678716bbd
📒 Files selected for processing (4)
pkg/cloud/gcp/actuators/machine/reconciler.gopkg/cloud/gcp/actuators/machine/reconciler_test.gopkg/cloud/gcp/actuators/services/compute/computeservice.gopkg/cloud/gcp/actuators/services/compute/computeservice_mock.go
|
@RadekManak can we get help getting this reviewed please - its a release blocker |
f5b6f13 to
73fee98
Compare
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
pkg/cloud/gcp/actuators/machine/reconciler_test.go (1)
197-203: AssertZoneOperationsListquery arguments to lock behavior.These mocks drive return shape, but they don’t validate the
filterandorderByinputs. Adding assertions for expected query content would prevent silent regressions inlatestVisibleInsertOperation()construction.Also applies to: 220-224, 1178-1187
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@pkg/cloud/gcp/actuators/machine/reconciler_test.go` around lines 197 - 203, The mock implementation for mockZoneOperationsList used in tests should assert the incoming query args (filter and orderBy, and optionally project/zone) to ensure latestVisibleInsertOperation() constructs the correct query; update the mock callbacks (mockZoneOperationsList at the instances around the given diffs and the ones at lines ~220 and ~1178) to check that the received filter string contains the expected operation type and instance metadata and that orderBy equals the expected sort (e.g., "insertTime desc") and fail the test (return an error or call t.Fatalf) when they don’t match so regressions in query construction are detected.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@pkg/cloud/gcp/actuators/machine/reconciler.go`:
- Around line 742-749: The loop over op.Error.Errors panics if any slice element
is nil — modify the block in reconciler.go that iterates "for _, e := range
op.Error.Errors" to skip nil entries (if e == nil continue) and also defensively
ensure op.Error and op.Error.Errors are non-nil before iterating; then build
msgs using e.Code and e.Message as before and return the formatted error string.
This prevents a nil dereference while preserving the existing error-message
aggregation.
---
Nitpick comments:
In `@pkg/cloud/gcp/actuators/machine/reconciler_test.go`:
- Around line 197-203: The mock implementation for mockZoneOperationsList used
in tests should assert the incoming query args (filter and orderBy, and
optionally project/zone) to ensure latestVisibleInsertOperation() constructs the
correct query; update the mock callbacks (mockZoneOperationsList at the
instances around the given diffs and the ones at lines ~220 and ~1178) to check
that the received filter string contains the expected operation type and
instance metadata and that orderBy equals the expected sort (e.g., "insertTime
desc") and fail the test (return an error or call t.Fatalf) when they don’t
match so regressions in query construction are detected.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository: openshift/coderabbit/.coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 35e9b6bd-5ef3-4114-a95c-fbccf95ed61d
📒 Files selected for processing (4)
pkg/cloud/gcp/actuators/machine/reconciler.gopkg/cloud/gcp/actuators/machine/reconciler_test.gopkg/cloud/gcp/actuators/services/compute/computeservice.gopkg/cloud/gcp/actuators/services/compute/computeservice_mock.go
73fee98 to
3326eaf
Compare
|
Note GitHub couldn't provide a complete incremental comparison for this pull request, so CodeRabbit is performing a full review instead. This review may take a little longer. |
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (3)
pkg/cloud/gcp/actuators/machine/reconciler_test.go (3)
68-69: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winBuild a fresh operation per test case instead of sharing
asyncFailureOp.
asyncFailureOpis created once and shared by allTestCreatecases. The case at Line 233 callswithTargetLink(asyncFailureOp, ...), which mutates that shared pointer in place from inside a mock closure. The case at Line 168 returns the same pointer fromMockInstancesInsert.The cases run sequentially today, so the assertions still pass. The coupling is fragile: adding
t.Parallel()or reordering the cases changes which subtest observes a mutatedTargetLink. CallzoneResourcePoolExhaustedOperationinside each mock, or makewithTargetLinkcopy the operation.♻️ Proposed change to make `withTargetLink` non-mutating
func withTargetLink(op *compute.Operation, targetLink string) *compute.Operation { - op.TargetLink = targetLink - return op + copied := *op + copied.TargetLink = targetLink + return &copied }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@pkg/cloud/gcp/actuators/machine/reconciler_test.go` around lines 68 - 69, Stop sharing the mutable asyncFailureOp across TestCreate cases: create a fresh operation with zoneResourcePoolExhaustedOperation inside each mock that returns or modifies it, including the withTargetLink path, or update withTargetLink to copy before changing TargetLink. Keep each subtest isolated so MockInstancesInsert and related assertions never observe another case’s mutation.
1440-1442: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueUse
errors.Asfor the requeue type check.Line 1440 asserts the concrete type directly. Line 1336 in the same file uses
errors.As. Useerrors.Ashere too, so a wrapped*machinecontroller.RequeueAfterErroris still detected.♻️ Proposed change
- if _, ok := err.(*machinecontroller.RequeueAfterError); ok { + var requeueErr *machinecontroller.RequeueAfterError + if errors.As(err, &requeueErr) { t.Fatalf("expected non-requeue error, got %v", err) }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@pkg/cloud/gcp/actuators/machine/reconciler_test.go` around lines 1440 - 1442, Update the error-type assertion in the test around the requeue check to use errors.As, matching the existing pattern near the other requeue assertion. Ensure wrapped *machinecontroller.RequeueAfterError values are detected while preserving the current failure message and non-requeue expectation.
171-190: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueMerge the two identical requeue cases.
"Requeue when returned operation is still RUNNING" and "Requeue when returned operation is not visible yet" use the same mocks and expect the same error. Only the operation name differs, and the reconciler does not read that name. Keep one case.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@pkg/cloud/gcp/actuators/machine/reconciler_test.go` around lines 171 - 190, Merge the duplicate requeue cases in reconciler_test.go by keeping a single table entry around the existing requeue path in the machine reconciler test. The two cases that cover a RUNNING operation and an operation not yet visible use the same mockInstancesInsert and mockInstancesGet behavior and assert the same machinecontroller.RequeueAfterError, so remove one of them and preserve the shared expectation in the remaining case.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@pkg/cloud/gcp/actuators/machine/reconciler.go`:
- Around line 726-733: The filter loop in the operation lookup should not return
immediately when ZoneOperationsList fails. Update the logic around
ZoneOperationsList to record the latest error and continue trying subsequent
filters, returning the operation result on success and the last recorded error
only after every filter has failed.
---
Nitpick comments:
In `@pkg/cloud/gcp/actuators/machine/reconciler_test.go`:
- Around line 68-69: Stop sharing the mutable asyncFailureOp across TestCreate
cases: create a fresh operation with zoneResourcePoolExhaustedOperation inside
each mock that returns or modifies it, including the withTargetLink path, or
update withTargetLink to copy before changing TargetLink. Keep each subtest
isolated so MockInstancesInsert and related assertions never observe another
case’s mutation.
- Around line 1440-1442: Update the error-type assertion in the test around the
requeue check to use errors.As, matching the existing pattern near the other
requeue assertion. Ensure wrapped *machinecontroller.RequeueAfterError values
are detected while preserving the current failure message and non-requeue
expectation.
- Around line 171-190: Merge the duplicate requeue cases in reconciler_test.go
by keeping a single table entry around the existing requeue path in the machine
reconciler test. The two cases that cover a RUNNING operation and an operation
not yet visible use the same mockInstancesInsert and mockInstancesGet behavior
and assert the same machinecontroller.RequeueAfterError, so remove one of them
and preserve the shared expectation in the remaining case.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository: openshift/coderabbit/.coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 0042d4f6-7a9f-4de6-b6b2-0a4f59647a8e
📒 Files selected for processing (4)
pkg/cloud/gcp/actuators/machine/reconciler.gopkg/cloud/gcp/actuators/machine/reconciler_test.gopkg/cloud/gcp/actuators/services/compute/computeservice.gopkg/cloud/gcp/actuators/services/compute/computeservice_mock.go
🚧 Files skipped from review as they are similar to previous changes (2)
- pkg/cloud/gcp/actuators/services/compute/computeservice.go
- pkg/cloud/gcp/actuators/services/compute/computeservice_mock.go
| for _, filter := range []string{ | ||
| fmt.Sprintf(`targetLink="%s" AND operationType="insert"`, fmtInstanceSelfLink(r.projectID, r.providerSpec.Zone, r.machine.Name)), | ||
| fmt.Sprintf(`operationType="insert" AND targetLink:"%s"`, expectedPath), | ||
| } { | ||
| opList, err := r.computeService.ZoneOperationsList(r.projectID, r.providerSpec.Zone, filter, "creationTimestamp desc") | ||
| if err != nil { | ||
| return nil, err | ||
| } |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟡 Minor | ⚡ Quick win
🧩 Analysis chain
🌐 Web query:
GCP Compute Engine API list filter syntax AND operator colon substring match targetLink zoneOperations
💡 Result:
The Google Cloud Compute Engine API supports filtering list requests using a syntax based on API Improvement Proposal AIP-160 [1][2][3]. Filter Syntax Details: Operators: The API supports the following operators: =,!=, >, <, <=, >=, and: [1][2][3]. AND/OR Logic: Expressions are combined using implicit AND logic when multiple separate expressions are provided within parentheses [1][2][3]. You can also explicitly use the AND and OR operators [1][2][3]. Substring Matching: The colon (:) operator is used for substring matching on string fields [3]. For non-string fields, the: operator functions identically to the = (equality) operator [3]. TargetLink and zoneOperations: When filtering zoneOperations, the targetLink field is commonly used to narrow results to specific resources [4]. Because the: operator performs a substring match, filtering by targetLink often requires extra verification in your code to ensure exact matches, as the filter may return any operation containing that string within the targetLink URL [4]. Example usage for zoneOperations: To filter for specific preempted instances, you might use a filter string such as: 'operationType="compute.instances.preempted" AND targetLink:instances/your-instance-name' [4] Note that because targetLink:instances/your-instance-name matches any targetLink containing that substring, it is recommended to further validate the results in your client-side code if strict matching is required [4].
Citations:
- 1: https://docs.cloud.google.com/compute/docs/reference/rest/v1/instanceGroups/listInstances
- 2: https://docs.cloud.google.com/python/docs/reference/compute/latest/google.cloud.compute_v1.types.ListInstancesRequest
- 3: https://googleapis.dev/nodejs/compute/latest/v1.InstancesClient.html
- 4: https://docs.cloud.google.com/compute/docs/samples/compute-preemptible-history
Don't return early on API errors; try the fallback filter.
Both filter operators are valid in the Compute API: AND combines expressions and : performs substring matching on string fields. However, if ZoneOperationsList returns an error on the first filter request—whether due to transient API issues or other reasons—the loop exits immediately. The fallback filter never runs. This causes the async-failure surfacing feature added by this PR to stop working silently. Continue to the next filter on error and return the last error only when all filters fail.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@pkg/cloud/gcp/actuators/machine/reconciler.go` around lines 726 - 733, The
filter loop in the operation lookup should not return immediately when
ZoneOperationsList fails. Update the logic around ZoneOperationsList to record
the latest error and continue trying subsequent filters, returning the operation
result on success and the last recorded error only after every filter has
failed.
ca1f83b to
7b00b61
Compare
Make provider-side instance creation failures visible during reconciliation by inspecting the returned insert operation and the latest visible zone insert operation, instead of treating a successful InstancesInsert HTTP response (or a missing instance alone) as enough. On failed DONE ops (e.g. ZONE_RESOURCE_POOL_EXHAUSTED), record MachineCreated=False with the provider error and retry create until it succeeds. Requeue while an insert is still RUNNING; on 409, reconcile (requeue on transient 404) rather than failing as InvalidMachineConfiguration. If a prior insert DONE successfully, skip re-insert when the instance exists, or recreate when it does not. Match operation targetLink by resource path so www/compute/sovereign-cloud URL hosts still resolve.
7b00b61 to
72afe38
Compare
…y stockout Extend async insert-op surfacing so providerID, addresses, and MachineCreated=True are set only when the instance is RUNNING with a clean or absent insert op. Capacity pool-exhausted failures keep MachineCreated=False with linear backoff and re-insert from Create after Exists=false, failing terminal after 30 minutes.
…r backoff MAO ignores InvalidMachineConfiguration from Update, so surface insert-op failures while the instance is visible via MachineCreated=False + requeue only. Terminal Failed stays on Create after vanish. Capacity backoff uses triangular 20s/40s/60s… waits; insert gate shares that schedule.
…rt failures Create sites that hit capacity via sync InstancesInsert or an immediate insert op error now share deadline logic with capacityInsertGate so they return InvalidMachineConfiguration past 30m instead of requeueing forever when no prior Operation is discoverable. Update still uses recordCapacityAndRequeue and never surfaces IMC.
Structure-only cleanup: collapse Create/Update capacity wrappers into evaluateCapacityRetry, share the triangular boundary walker, and merge duplicate insert-failure handling without changing behavior.
…acity ZoneOperations.list ignores targetLink HAS/substring (0 items live), so capacity/quota surfacing never ran. Match full targetLink URLs (www + compute + BasePath) and order by insertTime. Treat QUOTA_EXCEEDED like pool-exhausted with the same backoff/deadline, and gate Create/Update from MachineCreated=False when the op list still misses.
|
/hold |
openshift-ci
Bot
added
the
do-not-merge/hold
|
@RadekManak: The following test failed, say
Full PR test history. Your PR dashboard. DetailsInstructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here. |
3 participants
Make provider-side instance creation failures visible during reconciliation by checking the returned insert operation and visible in-flight operations instead of relying only on instance lookup.
Summary by CodeRabbit
New Features
Bug Fixes
Tests