-
Notifications
You must be signed in to change notification settings - Fork 485
/
events.go
266 lines (225 loc) · 7.15 KB
/
events.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
package rules
import (
"context"
"fmt"
"regexp"
"time"
"github.com/go-kit/log"
"github.com/grafana/agent/pkg/flow/logging/level"
"github.com/hashicorp/go-multierror"
promv1 "github.com/prometheus-operator/prometheus-operator/pkg/apis/monitoring/v1"
"github.com/prometheus/prometheus/model/rulefmt"
"k8s.io/client-go/tools/cache"
"k8s.io/client-go/util/workqueue"
"sigs.k8s.io/yaml" // Used for CRD compatibility instead of gopkg.in/yaml.v2
)
// This type must be hashable, so it is kept simple. The indexer will maintain a
// cache of current state, so this is mostly used for logging.
type event struct {
typ eventType
objectKey string
}
type eventType string
const (
eventTypeResourceChanged eventType = "resource-changed"
eventTypeSyncMimir eventType = "sync-mimir"
)
type queuedEventHandler struct {
log log.Logger
queue workqueue.RateLimitingInterface
}
func newQueuedEventHandler(log log.Logger, queue workqueue.RateLimitingInterface) *queuedEventHandler {
return &queuedEventHandler{
log: log,
queue: queue,
}
}
// OnAdd implements the cache.ResourceEventHandler interface.
func (c *queuedEventHandler) OnAdd(obj interface{}, _ bool) {
c.publishEvent(obj)
}
// OnUpdate implements the cache.ResourceEventHandler interface.
func (c *queuedEventHandler) OnUpdate(oldObj, newObj interface{}) {
c.publishEvent(newObj)
}
// OnDelete implements the cache.ResourceEventHandler interface.
func (c *queuedEventHandler) OnDelete(obj interface{}) {
c.publishEvent(obj)
}
func (c *queuedEventHandler) publishEvent(obj interface{}) {
key, err := cache.MetaNamespaceKeyFunc(obj)
if err != nil {
level.Error(c.log).Log("msg", "failed to get key for object", "err", err)
return
}
c.queue.AddRateLimited(event{
typ: eventTypeResourceChanged,
objectKey: key,
})
}
func (c *Component) eventLoop(ctx context.Context) {
for {
eventInterface, shutdown := c.queue.Get()
if shutdown {
level.Info(c.log).Log("msg", "shutting down event loop")
return
}
evt := eventInterface.(event)
c.metrics.eventsTotal.WithLabelValues(string(evt.typ)).Inc()
err := c.processEvent(ctx, evt)
if err != nil {
retries := c.queue.NumRequeues(evt)
if retries < 5 {
c.metrics.eventsRetried.WithLabelValues(string(evt.typ)).Inc()
c.queue.AddRateLimited(evt)
level.Error(c.log).Log(
"msg", "failed to process event, will retry",
"retries", fmt.Sprintf("%d/5", retries),
"err", err,
)
continue
} else {
c.metrics.eventsFailed.WithLabelValues(string(evt.typ)).Inc()
level.Error(c.log).Log(
"msg", "failed to process event, max retries exceeded",
"retries", fmt.Sprintf("%d/5", retries),
"err", err,
)
c.reportUnhealthy(err)
}
} else {
c.reportHealthy()
}
c.queue.Forget(evt)
}
}
func (c *Component) processEvent(ctx context.Context, e event) error {
defer c.queue.Done(e)
switch e.typ {
case eventTypeResourceChanged:
level.Info(c.log).Log("msg", "processing event", "type", e.typ, "key", e.objectKey)
case eventTypeSyncMimir:
level.Debug(c.log).Log("msg", "syncing current state from ruler")
err := c.syncMimir(ctx)
if err != nil {
return err
}
default:
return fmt.Errorf("unknown event type: %s", e.typ)
}
return c.reconcileState(ctx)
}
func (c *Component) syncMimir(ctx context.Context) error {
rulesByNamespace, err := c.mimirClient.ListRules(ctx, "")
if err != nil {
level.Error(c.log).Log("msg", "failed to list rules from mimir", "err", err)
return err
}
for ns := range rulesByNamespace {
if !isManagedMimirNamespace(c.args.MimirNameSpacePrefix, ns) {
delete(rulesByNamespace, ns)
}
}
c.currentState = rulesByNamespace
return nil
}
func (c *Component) reconcileState(ctx context.Context) error {
ctx, cancel := context.WithTimeout(ctx, 5*time.Second)
defer cancel()
desiredState, err := c.loadStateFromK8s()
if err != nil {
return err
}
diffs := diffRuleState(desiredState, c.currentState)
var result error
for ns, diff := range diffs {
err = c.applyChanges(ctx, ns, diff)
if err != nil {
result = multierror.Append(result, err)
continue
}
}
return result
}
func (c *Component) loadStateFromK8s() (ruleGroupsByNamespace, error) {
matchedNamespaces, err := c.namespaceLister.List(c.namespaceSelector)
if err != nil {
return nil, fmt.Errorf("failed to list namespaces: %w", err)
}
desiredState := make(ruleGroupsByNamespace)
for _, ns := range matchedNamespaces {
crdState, err := c.ruleLister.PrometheusRules(ns.Name).List(c.ruleSelector)
if err != nil {
return nil, fmt.Errorf("failed to list rules: %w", err)
}
for _, pr := range crdState {
mimirNs := mimirNamespaceForRuleCRD(c.args.MimirNameSpacePrefix, pr)
groups, err := convertCRDRuleGroupToRuleGroup(pr.Spec)
if err != nil {
return nil, fmt.Errorf("failed to convert rule group: %w", err)
}
desiredState[mimirNs] = groups
}
}
return desiredState, nil
}
func convertCRDRuleGroupToRuleGroup(crd promv1.PrometheusRuleSpec) ([]rulefmt.RuleGroup, error) {
buf, err := yaml.Marshal(crd)
if err != nil {
return nil, err
}
groups, errs := rulefmt.Parse(buf)
if len(errs) > 0 {
return nil, multierror.Append(nil, errs...)
}
return groups.Groups, nil
}
func (c *Component) applyChanges(ctx context.Context, namespace string, diffs []ruleGroupDiff) error {
if len(diffs) == 0 {
return nil
}
for _, diff := range diffs {
switch diff.Kind {
case ruleGroupDiffKindAdd:
err := c.mimirClient.CreateRuleGroup(ctx, namespace, diff.Desired)
if err != nil {
return err
}
level.Info(c.log).Log("msg", "added rule group", "namespace", namespace, "group", diff.Desired.Name)
case ruleGroupDiffKindRemove:
err := c.mimirClient.DeleteRuleGroup(ctx, namespace, diff.Actual.Name)
if err != nil {
return err
}
level.Info(c.log).Log("msg", "removed rule group", "namespace", namespace, "group", diff.Actual.Name)
case ruleGroupDiffKindUpdate:
err := c.mimirClient.CreateRuleGroup(ctx, namespace, diff.Desired)
if err != nil {
return err
}
level.Info(c.log).Log("msg", "updated rule group", "namespace", namespace, "group", diff.Desired.Name)
default:
level.Error(c.log).Log("msg", "unknown rule group diff kind", "kind", diff.Kind)
}
}
// resync mimir state after applying changes
return c.syncMimir(ctx)
}
// mimirNamespaceForRuleCRD returns the namespace that the rule CRD should be
// stored in mimir. This function, along with isManagedNamespace, is used to
// determine if a rule CRD is managed by the agent.
func mimirNamespaceForRuleCRD(prefix string, pr *promv1.PrometheusRule) string {
return fmt.Sprintf("%s/%s/%s/%s", prefix, pr.Namespace, pr.Name, pr.UID)
}
// isManagedMimirNamespace returns true if the namespace is managed by the agent.
// Unmanaged namespaces are left as is by the operator.
func isManagedMimirNamespace(prefix, namespace string) bool {
prefixPart := regexp.QuoteMeta(prefix)
namespacePart := `.+`
namePart := `.+`
uuidPart := `[0-9a-fA-F]{8}\b-[0-9a-fA-F]{4}\b-[0-9a-fA-F]{4}\b-[0-9a-fA-F]{4}\b-[0-9a-fA-F]{12}`
managedNamespaceRegex := regexp.MustCompile(
fmt.Sprintf("^%s/%s/%s/%s$", prefixPart, namespacePart, namePart, uuidPart),
)
return managedNamespaceRegex.MatchString(namespace)
}