Replace http.DefaultClient with a package-level *http.Client with an
explicit 10-second timeout. Avoids sharing the process-wide default
transport with other libraries.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Pass bucket names from BucketClaim.Status.BucketName into the PromQL
query as a bucket=~"name1|name2" filter. This prevents O(N²) load
where N WorkloadMonitors each fetch all buckets globally.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Wrap resolvePrometheusURL and queryAllBucketMetrics in a
len(bucketClaimList.Items) > 0 guard. Avoids unnecessary namespace
GET and HTTP request to Prometheus on every reconcile of non-bucket
WorkloadMonitors (postgres, redis, kubernetes, etc.).
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Replace 2×N per-bucket HTTP requests with a single query that fetches
all SeaweedFS bucket size metrics at once:
{__name__=~"SeaweedFS_s3_bucket_(size|physical_size)_bytes"}
Results are keyed by bucket name in memory, then looked up per
BucketClaim. This reduces HTTP round-trips from 2N+1 to 2 (one
namespace lookup + one Prometheus query) regardless of bucket count.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Move resolvePrometheusURL call before the BucketClaim loop and pass
the URL as parameter. Avoids redundant namespace lookups and reduces
reconcile time from 2×N+1 to 2×N HTTP calls (where N = BucketClaims).
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Change queryPrometheusMetric to return (int64, bool) so callers can
emit s3-storage-bytes=0 for empty buckets while omitting the field
entirely when monitoring is not configured.
Also:
- Add RBAC marker for core/namespaces GET (used by resolvePrometheusURL)
- Log namespace read errors instead of silently returning empty URL
- Add TestQueryPrometheusMetricZeroValue test
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Bucket existence is already tracked via LifetimeHours. The s3-buckets
count resource produced a meaningless "s3-buckets-Hours" fallback type
in billing. Only storage size metrics (s3-storage-bytes,
s3-physical-storage-bytes) are now set on bucket Workloads.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
- Initialize workload.Labels map inside CreateOrUpdate mutate function
to prevent nil map panic when existing Workload has no labels
- Use strconv.ParseFloat instead of resource.ParseQuantity for
Prometheus metric values, which may use scientific notation
(e.g. "1.048576e+06") that ParseQuantity does not support
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Replace static --prometheus-url flag with dynamic resolution from
namespace.cozystack.io/monitoring label. Each tenant namespace knows
which tenant hosts its monitoring stack, so the controller constructs
the vmselect URL automatically. This correctly handles multi-tenant
setups where different tenants may use different monitoring instances.
- Remove PrometheusURL field from WorkloadMonitorReconciler struct
- Remove --prometheus-url flag and prometheusUrl chart value
- Add resolvePrometheusURL() that reads namespace label
- queryPrometheusMetric() now accepts prometheusBaseURL as parameter
- Add tests for resolvePrometheusURL with and without label
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
- Set replicas/minReplicas to 0 in bucket WorkloadMonitor (buckets
have no pods, minReplicas=1 would mark monitor as not operational)
- Remove dead IsNotFound check on List (List returns empty, not 404)
- Trim trailing slash from PrometheusURL before path append
- Add strings import for TrimRight
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
- Check resp.StatusCode before parsing Prometheus response
- Limit response body read to 1 MB via io.LimitReader
- Use strings.HasPrefix in test instead of fragile slice indexing
- Add TestQueryPrometheusMetricServerError for 500 responses
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Query SeaweedFS_s3_bucket_physical_size_bytes alongside the logical
size metric. Physical size includes all replicas and reflects actual
disk usage, while logical size reflects what the user stored.
Refactor queryBucketSizeBytes into generic queryPrometheusMetric
to reuse for both metrics.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Query SeaweedFS_s3_bucket_size_bytes from a Prometheus-compatible API
to populate s3-storage-bytes resource on bucket Workloads. The
Prometheus URL is configurable via --prometheus-url flag. When set,
bucket WorkloadMonitors are requeued every 60s to keep sizes current.
When Prometheus is not configured, buckets still get tracked with
s3-buckets=1 for existence-based billing.
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Add reconcileBucketClaimForMonitor() that watches COSI BucketClaim
objects and creates Workload CRDs with s3-buckets resource, following
the same pattern as PVC and Service reconcilers. This enables the
billing pipeline to discover and track S3 buckets per tenant.
Changes:
- Add COSI API types dependency (container-object-storage-interface-api)
- Register cosiv1alpha1 scheme in controller main
- Add BucketClaim watch in SetupWithManager
- Add BucketClaim list + reconcile in Reconcile loop
- Add RBAC annotation for objectstorage.k8s.io/bucketclaims
- Add unit tests for BucketClaim reconciliation
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: ZverGuy <maximbel2003@gmail.com>
Inside the RetryOnConflict block, derive the operational status from
fresh.Spec.MinReplicas instead of the stale monitor.Spec.MinReplicas
so that concurrent spec updates observed by the retry are respected.
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
Add two features to improve application observability in the dashboard:
1. WorkloadsReady condition on Application status
- Query WorkloadMonitor resources to determine if all pods are running
- Expose WorkloadsReady as a separate condition alongside Ready
- Ready continues to reflect HelmRelease state only (no override) to
preserve backward compatibility with tooling and avoid false-negatives
during normal startup windows when pods are still coming up
- Handle three states: operational, not operational, unknown (pending)
- Fail-open on WorkloadMonitor query errors
- Integrate with Application Watch to emit MODIFIED events on
WorkloadMonitor changes (with initial-events-end safety)
- Register cozystack.io/v1alpha1 types in API server scheme
2. Events tab in dashboard
- Show Kubernetes Events scoped to application namespace
- Use status.namespace for Tenant applications
- Include both lastTimestamp and eventTime columns for K8s compat
3. Bug fix: WorkloadMonitor Operational status persistence
- Operational was written to stale 'monitor' instead of 'fresh'
inside RetryOnConflict, so it was never persisted to the cluster
Closes#2359Closes#2360
Assisted-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
When the WorkloadMonitor is reconciled and child Workload objects are
created, they will now get additional labels in the
`workloads.cozystack.io` namespace, containing metadata about the
workload. This particular commit checks if a pod targeted by a Workload
is owned by a VirtualMachineInstance (i.e. it launches a KubeVirt VMI)
and, if so, gets the VMI instance type and puts it in the
`kubevirt-vmi-instance-type` label.
Signed-off-by: Timofei Larkin <lllamnyp@gmail.com>
* Count Workload resources for pods by requests, not limits
* Do not count init container requests
* Prefix Workloads for pods with `pod-`, just like the other types to
prevent possible name collisions (closes#787)
The previous version of the WorkloadMonitor controller incorrectly
summed resource limits on pods, rather than requests. This prevented it
from tracking the resource allocation for pods, which only had requests
specified, which is particularly the case for kubevirt's virtual machine
pods. Additionally, it counted the limits for all containers, including
init containers, which are short-lived and do not contribute much to the
total resource usage.
Signed-off-by: Timofei Larkin <lllamnyp@gmail.com>
When populating the WorkloadMonitor objects, the status field is now
populated with a specially formatted string, mimicking the keys of
ResourceQuota.spec.hard, e.g.
`<storageclassname>.storageclass.storage.k8s.io/requests.storage` or
`<ipaddresspoolname>.ipaddresspool.metallb.io/requests.ipaddresses`
so the storage class or IP pool in use can be tracked. Part of #788.
Signed-off-by: Timofei Larkin <lllamnyp@gmail.com>
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
Based on the comprehensive summary of changes, here are the release
notes:
- **New Features**
- Added a new Kubernetes controller for managing workload monitoring
- Introduced telemetry collection capabilities with configurable options
- Added new Custom Resource Definitions (CRDs) for Workload and
WorkloadMonitor
- **Improvements**
- Enhanced API infrastructure with new API group and version
- Improved deployment configurations for various system components
- Added development container and workflow configurations
- **Bug Fixes**
- Updated import paths to correct domain naming
- **Chores**
- Updated copyright years
- Refined module dependencies
- Standardized code linting and testing configurations
- **Infrastructure**
- Increased `cozystack-api` deployment replicas from 1 to 2 for improved
availability
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Signed-off-by: Andrei Kvapil <kvapss@gmail.com>
