2015-01-21 19:07:45 +00:00
|
|
|
// Copyright 2013 The Prometheus Authors
|
2013-01-07 22:24:26 +00:00
|
|
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
|
|
|
// you may not use this file except in compliance with the License.
|
|
|
|
// You may obtain a copy of the License at
|
|
|
|
//
|
|
|
|
// http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
//
|
|
|
|
// Unless required by applicable law or agreed to in writing, software
|
|
|
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
|
|
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
|
|
// See the License for the specific language governing permissions and
|
|
|
|
// limitations under the License.
|
|
|
|
|
2015-03-30 17:43:19 +00:00
|
|
|
package rules
|
2013-01-07 22:24:26 +00:00
|
|
|
|
|
|
|
import (
|
2013-12-10 15:54:35 +00:00
|
|
|
"fmt"
|
2015-04-29 09:08:56 +00:00
|
|
|
"io/ioutil"
|
2017-05-18 15:26:36 +00:00
|
|
|
"math"
|
2015-06-30 12:38:01 +00:00
|
|
|
"net/url"
|
2015-05-28 05:36:21 +00:00
|
|
|
"path/filepath"
|
2013-04-17 12:42:15 +00:00
|
|
|
"sync"
|
2013-01-07 22:24:26 +00:00
|
|
|
"time"
|
|
|
|
|
2015-06-30 09:51:05 +00:00
|
|
|
html_template "html/template"
|
|
|
|
|
2014-06-18 17:43:15 +00:00
|
|
|
"github.com/prometheus/client_golang/prometheus"
|
2015-10-03 08:21:43 +00:00
|
|
|
"github.com/prometheus/common/log"
|
2015-08-20 15:18:46 +00:00
|
|
|
"github.com/prometheus/common/model"
|
2016-09-15 22:58:06 +00:00
|
|
|
"golang.org/x/net/context"
|
2013-06-25 12:02:27 +00:00
|
|
|
|
2015-05-12 14:52:56 +00:00
|
|
|
"github.com/prometheus/prometheus/config"
|
2016-03-01 11:37:22 +00:00
|
|
|
"github.com/prometheus/prometheus/notifier"
|
2017-05-18 15:26:36 +00:00
|
|
|
"github.com/prometheus/prometheus/pkg/labels"
|
|
|
|
"github.com/prometheus/prometheus/pkg/timestamp"
|
|
|
|
"github.com/prometheus/prometheus/pkg/value"
|
2015-03-30 17:43:19 +00:00
|
|
|
"github.com/prometheus/prometheus/promql"
|
2015-03-15 02:36:15 +00:00
|
|
|
"github.com/prometheus/prometheus/storage"
|
2015-05-29 11:30:30 +00:00
|
|
|
"github.com/prometheus/prometheus/util/strutil"
|
2013-06-25 12:02:27 +00:00
|
|
|
)
|
2013-01-07 22:24:26 +00:00
|
|
|
|
2014-06-18 17:43:15 +00:00
|
|
|
// Constants for instrumentation.
|
2016-01-21 10:09:24 +00:00
|
|
|
const namespace = "prometheus"
|
2014-06-18 17:43:15 +00:00
|
|
|
|
|
|
|
var (
|
|
|
|
evalDuration = prometheus.NewSummaryVec(
|
|
|
|
prometheus.SummaryOpts{
|
2014-07-23 17:55:33 +00:00
|
|
|
Namespace: namespace,
|
2015-12-15 12:15:07 +00:00
|
|
|
Name: "rule_evaluation_duration_seconds",
|
2014-07-23 17:55:33 +00:00
|
|
|
Help: "The duration for a rule to execute.",
|
2014-06-18 17:43:15 +00:00
|
|
|
},
|
2016-01-21 10:09:24 +00:00
|
|
|
[]string{"rule_type"},
|
2014-06-18 17:43:15 +00:00
|
|
|
)
|
2016-01-21 10:09:24 +00:00
|
|
|
evalFailures = prometheus.NewCounterVec(
|
2014-12-31 12:16:08 +00:00
|
|
|
prometheus.CounterOpts{
|
|
|
|
Namespace: namespace,
|
|
|
|
Name: "rule_evaluation_failures_total",
|
|
|
|
Help: "The total number of rule evaluation failures.",
|
|
|
|
},
|
2016-01-21 10:09:24 +00:00
|
|
|
[]string{"rule_type"},
|
2014-12-31 12:16:08 +00:00
|
|
|
)
|
2016-01-21 10:09:24 +00:00
|
|
|
evalTotal = prometheus.NewCounterVec(
|
2015-12-15 12:15:07 +00:00
|
|
|
prometheus.CounterOpts{
|
|
|
|
Namespace: namespace,
|
|
|
|
Name: "rule_evaluations_total",
|
|
|
|
Help: "The total number of rule evaluations.",
|
|
|
|
},
|
2016-01-21 10:09:24 +00:00
|
|
|
[]string{"rule_type"},
|
2015-12-15 12:15:07 +00:00
|
|
|
)
|
2014-07-23 17:55:33 +00:00
|
|
|
iterationDuration = prometheus.NewSummary(prometheus.SummaryOpts{
|
|
|
|
Namespace: namespace,
|
2015-12-15 12:15:07 +00:00
|
|
|
Name: "evaluator_duration_seconds",
|
2016-01-27 18:07:46 +00:00
|
|
|
Help: "The duration of rule group evaluations.",
|
2015-01-21 14:42:25 +00:00
|
|
|
Objectives: map[float64]float64{0.01: 0.001, 0.05: 0.005, 0.5: 0.05, 0.90: 0.01, 0.99: 0.001},
|
2014-07-23 17:55:33 +00:00
|
|
|
})
|
2016-01-27 18:07:46 +00:00
|
|
|
iterationsSkipped = prometheus.NewCounter(prometheus.CounterOpts{
|
|
|
|
Namespace: namespace,
|
|
|
|
Name: "evaluator_iterations_skipped_total",
|
|
|
|
Help: "The total number of rule group evaluations skipped due to throttled metric storage.",
|
|
|
|
})
|
2017-04-02 23:03:28 +00:00
|
|
|
iterationsMissed = prometheus.NewCounter(prometheus.CounterOpts{
|
|
|
|
Namespace: namespace,
|
|
|
|
Name: "evaluator_iterations_missed_total",
|
|
|
|
Help: "The total number of rule group evaluations missed due to slow rule group evaluation.",
|
|
|
|
})
|
2016-01-27 18:07:46 +00:00
|
|
|
iterationsScheduled = prometheus.NewCounter(prometheus.CounterOpts{
|
|
|
|
Namespace: namespace,
|
|
|
|
Name: "evaluator_iterations_total",
|
2017-04-02 23:03:28 +00:00
|
|
|
Help: "The total number of scheduled rule group evaluations, whether executed, missed or skipped.",
|
2016-01-27 18:07:46 +00:00
|
|
|
})
|
2014-06-18 17:43:15 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
func init() {
|
2016-01-21 10:09:24 +00:00
|
|
|
evalTotal.WithLabelValues(string(ruleTypeAlert))
|
|
|
|
evalTotal.WithLabelValues(string(ruleTypeRecording))
|
|
|
|
evalFailures.WithLabelValues(string(ruleTypeAlert))
|
|
|
|
evalFailures.WithLabelValues(string(ruleTypeRecording))
|
|
|
|
|
2014-06-18 17:43:15 +00:00
|
|
|
prometheus.MustRegister(iterationDuration)
|
2017-04-02 23:32:56 +00:00
|
|
|
prometheus.MustRegister(iterationsScheduled)
|
2016-01-27 18:07:46 +00:00
|
|
|
prometheus.MustRegister(iterationsSkipped)
|
2017-04-02 23:03:28 +00:00
|
|
|
prometheus.MustRegister(iterationsMissed)
|
2014-12-31 12:16:08 +00:00
|
|
|
prometheus.MustRegister(evalFailures)
|
2014-06-18 17:43:15 +00:00
|
|
|
prometheus.MustRegister(evalDuration)
|
|
|
|
}
|
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
type ruleType string
|
|
|
|
|
|
|
|
const (
|
|
|
|
ruleTypeAlert = "alerting"
|
|
|
|
ruleTypeRecording = "recording"
|
|
|
|
)
|
|
|
|
|
2015-06-30 09:18:07 +00:00
|
|
|
// A Rule encapsulates a vector expression which is evaluated at a specified
|
|
|
|
// interval and acted upon (currently either recorded or used for alerting).
|
|
|
|
type Rule interface {
|
|
|
|
Name() string
|
2015-12-14 16:40:40 +00:00
|
|
|
// eval evaluates the rule, including any associated recording or alerting actions.
|
2017-05-16 14:48:37 +00:00
|
|
|
Eval(context.Context, time.Time, *promql.Engine, *url.URL) (promql.Vector, error)
|
2015-06-30 09:18:07 +00:00
|
|
|
// String returns a human-readable string representation of the rule.
|
|
|
|
String() string
|
|
|
|
// HTMLSnippet returns a human-readable string representation of the rule,
|
|
|
|
// decorated with HTML elements for use the web frontend.
|
2015-06-30 09:51:05 +00:00
|
|
|
HTMLSnippet(pathPrefix string) html_template.HTML
|
2015-06-30 09:18:07 +00:00
|
|
|
}
|
|
|
|
|
2015-12-17 10:46:10 +00:00
|
|
|
// Group is a set of rules that have a logical relation.
|
2015-12-14 16:40:40 +00:00
|
|
|
type Group struct {
|
2017-05-18 15:26:36 +00:00
|
|
|
name string
|
|
|
|
interval time.Duration
|
|
|
|
rules []Rule
|
|
|
|
opts *ManagerOptions
|
|
|
|
seriesInPreviousEval map[string]labels.Labels
|
2013-08-20 13:42:06 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
done chan struct{}
|
|
|
|
terminated chan struct{}
|
|
|
|
}
|
2013-08-20 13:42:06 +00:00
|
|
|
|
2016-11-18 16:25:58 +00:00
|
|
|
// NewGroup makes a new Group with the given name, options, and rules.
|
2016-11-18 16:32:30 +00:00
|
|
|
func NewGroup(name string, interval time.Duration, rules []Rule, opts *ManagerOptions) *Group {
|
2015-12-14 16:40:40 +00:00
|
|
|
return &Group{
|
2017-05-18 15:26:36 +00:00
|
|
|
name: name,
|
|
|
|
interval: interval,
|
|
|
|
rules: rules,
|
|
|
|
opts: opts,
|
|
|
|
seriesInPreviousEval: map[string]labels.Labels{},
|
|
|
|
done: make(chan struct{}),
|
|
|
|
terminated: make(chan struct{}),
|
2015-12-14 16:40:40 +00:00
|
|
|
}
|
2013-08-20 13:42:06 +00:00
|
|
|
}
|
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
func (g *Group) run() {
|
|
|
|
defer close(g.terminated)
|
2013-08-20 13:42:06 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
// Wait an initial amount to have consistently slotted intervals.
|
2016-01-12 09:52:40 +00:00
|
|
|
select {
|
|
|
|
case <-time.After(g.offset()):
|
|
|
|
case <-g.done:
|
|
|
|
return
|
|
|
|
}
|
2013-08-20 13:42:06 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
iter := func() {
|
2016-01-27 18:07:46 +00:00
|
|
|
iterationsScheduled.Inc()
|
2016-12-29 08:27:30 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
start := time.Now()
|
2017-05-18 16:47:00 +00:00
|
|
|
g.Eval(start)
|
2013-01-07 22:24:26 +00:00
|
|
|
|
2016-07-07 13:24:35 +00:00
|
|
|
iterationDuration.Observe(time.Since(start).Seconds())
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
2017-04-02 23:03:28 +00:00
|
|
|
lastTriggered := time.Now()
|
2015-12-14 16:40:40 +00:00
|
|
|
iter()
|
2015-01-29 14:05:10 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
tick := time.NewTicker(g.interval)
|
|
|
|
defer tick.Stop()
|
2013-01-07 22:24:26 +00:00
|
|
|
|
|
|
|
for {
|
|
|
|
select {
|
2015-12-14 16:40:40 +00:00
|
|
|
case <-g.done:
|
2013-12-11 14:30:27 +00:00
|
|
|
return
|
2015-01-29 14:05:10 +00:00
|
|
|
default:
|
|
|
|
select {
|
2015-12-14 16:40:40 +00:00
|
|
|
case <-g.done:
|
2015-01-29 14:05:10 +00:00
|
|
|
return
|
2015-12-14 16:40:40 +00:00
|
|
|
case <-tick.C:
|
2017-04-02 23:03:28 +00:00
|
|
|
missed := (time.Since(lastTriggered).Nanoseconds() / g.interval.Nanoseconds()) - 1
|
|
|
|
if missed > 0 {
|
|
|
|
iterationsMissed.Add(float64(missed))
|
|
|
|
iterationsScheduled.Add(float64(missed))
|
|
|
|
}
|
|
|
|
lastTriggered = time.Now()
|
2015-12-14 16:40:40 +00:00
|
|
|
iter()
|
2015-01-29 14:05:10 +00:00
|
|
|
}
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
func (g *Group) stop() {
|
|
|
|
close(g.done)
|
|
|
|
<-g.terminated
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
func (g *Group) fingerprint() model.Fingerprint {
|
|
|
|
l := model.LabelSet{"name": model.LabelValue(g.name)}
|
|
|
|
return l.Fingerprint()
|
|
|
|
}
|
|
|
|
|
2015-12-17 10:46:10 +00:00
|
|
|
// offset returns until the next consistently slotted evaluation interval.
|
2015-12-14 16:40:40 +00:00
|
|
|
func (g *Group) offset() time.Duration {
|
|
|
|
now := time.Now().UnixNano()
|
|
|
|
|
|
|
|
var (
|
|
|
|
base = now - (now % int64(g.interval))
|
|
|
|
offset = uint64(g.fingerprint()) % uint64(g.interval)
|
|
|
|
next = base + int64(offset)
|
|
|
|
)
|
|
|
|
|
|
|
|
if next < now {
|
|
|
|
next += int64(g.interval)
|
2013-07-30 15:18:07 +00:00
|
|
|
}
|
2015-12-14 16:40:40 +00:00
|
|
|
return time.Duration(next - now)
|
|
|
|
}
|
2013-07-30 15:18:07 +00:00
|
|
|
|
2017-05-18 15:26:36 +00:00
|
|
|
// copyState copies the alerting rule and staleness related state from the given group.
|
2015-12-14 16:40:40 +00:00
|
|
|
func (g *Group) copyState(from *Group) {
|
2017-05-18 15:26:36 +00:00
|
|
|
g.seriesInPreviousEval = from.seriesInPreviousEval
|
2015-12-14 16:40:40 +00:00
|
|
|
for _, fromRule := range from.rules {
|
|
|
|
far, ok := fromRule.(*AlertingRule)
|
|
|
|
if !ok {
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
for _, rule := range g.rules {
|
|
|
|
ar, ok := rule.(*AlertingRule)
|
|
|
|
if !ok {
|
|
|
|
continue
|
|
|
|
}
|
2016-03-02 10:54:37 +00:00
|
|
|
// TODO(fabxc): forbid same alert definitions that are not unique by
|
|
|
|
// at least on static label or alertname?
|
|
|
|
if far.equal(ar) {
|
|
|
|
for fp, a := range far.active {
|
|
|
|
ar.active[fp] = a
|
|
|
|
}
|
2015-12-14 16:40:40 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-01-21 10:09:24 +00:00
|
|
|
func typeForRule(r Rule) ruleType {
|
|
|
|
switch r.(type) {
|
|
|
|
case *AlertingRule:
|
|
|
|
return ruleTypeAlert
|
|
|
|
case *RecordingRule:
|
|
|
|
return ruleTypeRecording
|
|
|
|
}
|
|
|
|
panic(fmt.Errorf("unknown rule type: %T", r))
|
|
|
|
}
|
|
|
|
|
2016-11-18 16:25:58 +00:00
|
|
|
// Eval runs a single evaluation cycle in which all rules are evaluated in parallel.
|
2015-12-17 10:46:10 +00:00
|
|
|
// In the future a single group will be evaluated sequentially to properly handle
|
|
|
|
// rule dependency.
|
2017-05-18 16:47:00 +00:00
|
|
|
func (g *Group) Eval(ts time.Time) {
|
2015-12-14 16:40:40 +00:00
|
|
|
var (
|
2017-05-18 15:26:36 +00:00
|
|
|
wg sync.WaitGroup
|
|
|
|
mu sync.Mutex
|
|
|
|
seriesReturned = make(map[string]labels.Labels, len(g.seriesInPreviousEval))
|
2015-12-14 16:40:40 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
for _, rule := range g.rules {
|
2016-01-21 10:09:24 +00:00
|
|
|
rtyp := string(typeForRule(rule))
|
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
wg.Add(1)
|
|
|
|
// BUG(julius): Look at fixing thundering herd.
|
|
|
|
go func(rule Rule) {
|
|
|
|
defer wg.Done()
|
|
|
|
|
2016-01-21 10:09:24 +00:00
|
|
|
defer func(t time.Time) {
|
2016-07-07 13:24:35 +00:00
|
|
|
evalDuration.WithLabelValues(rtyp).Observe(time.Since(t).Seconds())
|
2016-01-21 10:09:24 +00:00
|
|
|
}(time.Now())
|
|
|
|
|
|
|
|
evalTotal.WithLabelValues(rtyp).Inc()
|
2015-12-14 16:40:40 +00:00
|
|
|
|
2017-05-18 16:47:00 +00:00
|
|
|
vector, err := rule.Eval(g.opts.Context, ts, g.opts.QueryEngine, g.opts.ExternalURL)
|
2015-12-14 16:40:40 +00:00
|
|
|
if err != nil {
|
2016-01-18 15:53:37 +00:00
|
|
|
// Canceled queries are intentional termination of queries. This normally
|
|
|
|
// happens on shutdown and thus we skip logging of any errors here.
|
|
|
|
if _, ok := err.(promql.ErrQueryCanceled); !ok {
|
|
|
|
log.Warnf("Error while evaluating rule %q: %s", rule, err)
|
|
|
|
}
|
2016-01-21 10:09:24 +00:00
|
|
|
evalFailures.WithLabelValues(rtyp).Inc()
|
|
|
|
return
|
2015-12-14 16:40:40 +00:00
|
|
|
}
|
|
|
|
|
2016-01-21 10:09:24 +00:00
|
|
|
if ar, ok := rule.(*AlertingRule); ok {
|
2016-12-29 16:31:14 +00:00
|
|
|
g.sendAlerts(ar)
|
2015-12-14 16:40:40 +00:00
|
|
|
}
|
2016-05-19 14:22:49 +00:00
|
|
|
var (
|
|
|
|
numOutOfOrder = 0
|
|
|
|
numDuplicates = 0
|
|
|
|
)
|
2016-12-24 23:37:46 +00:00
|
|
|
|
2017-01-13 13:48:01 +00:00
|
|
|
app, err := g.opts.Appendable.Appender()
|
|
|
|
if err != nil {
|
|
|
|
log.With("err", err).Warn("creating appender failed")
|
|
|
|
return
|
|
|
|
}
|
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
for _, s := range vector {
|
2017-02-01 14:59:37 +00:00
|
|
|
if _, err := app.Add(s.Metric, s.T, s.V); err != nil {
|
2016-05-19 14:22:49 +00:00
|
|
|
switch err {
|
2016-12-29 08:27:30 +00:00
|
|
|
case storage.ErrOutOfOrderSample:
|
2016-05-19 14:22:49 +00:00
|
|
|
numOutOfOrder++
|
2017-02-01 14:59:37 +00:00
|
|
|
log.With("sample", s).With("err", err).Debug("Rule evaluation result discarded")
|
2016-12-29 08:27:30 +00:00
|
|
|
case storage.ErrDuplicateSampleForTimestamp:
|
2016-05-19 14:22:49 +00:00
|
|
|
numDuplicates++
|
2017-02-01 14:59:37 +00:00
|
|
|
log.With("sample", s).With("err", err).Debug("Rule evaluation result discarded")
|
2016-05-19 14:22:49 +00:00
|
|
|
default:
|
2017-02-01 14:59:37 +00:00
|
|
|
log.With("sample", s).With("err", err).Warn("Rule evaluation result discarded")
|
2016-05-19 14:22:49 +00:00
|
|
|
}
|
2017-05-18 15:26:36 +00:00
|
|
|
} else {
|
|
|
|
mu.Lock()
|
|
|
|
seriesReturned[s.Metric.String()] = s.Metric
|
|
|
|
mu.Unlock()
|
2016-05-19 14:22:49 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
if numOutOfOrder > 0 {
|
|
|
|
log.With("numDropped", numOutOfOrder).Warn("Error on ingesting out-of-order result from rule evaluation")
|
|
|
|
}
|
|
|
|
if numDuplicates > 0 {
|
|
|
|
log.With("numDropped", numDuplicates).Warn("Error on ingesting results from rule evaluation with different value but same timestamp")
|
2015-12-14 16:40:40 +00:00
|
|
|
}
|
2017-01-13 13:48:01 +00:00
|
|
|
if err := app.Commit(); err != nil {
|
2016-12-29 08:27:30 +00:00
|
|
|
log.With("err", err).Warn("rule sample appending failed")
|
|
|
|
}
|
2015-12-14 16:40:40 +00:00
|
|
|
}(rule)
|
|
|
|
}
|
|
|
|
wg.Wait()
|
2017-05-18 15:26:36 +00:00
|
|
|
|
|
|
|
// TODO(bbrazil): This should apply per-rule.
|
|
|
|
app, err := g.opts.Appendable.Appender()
|
|
|
|
if err != nil {
|
|
|
|
log.With("err", err).Warn("creating appender failed")
|
|
|
|
return
|
|
|
|
}
|
|
|
|
for metric, lset := range g.seriesInPreviousEval {
|
|
|
|
if _, ok := seriesReturned[metric]; !ok {
|
|
|
|
// Series no longer exposed, mark it stale.
|
2017-05-18 16:47:00 +00:00
|
|
|
_, err = app.Add(lset, timestamp.FromTime(ts), math.Float64frombits(value.StaleNaN))
|
2017-05-18 15:26:36 +00:00
|
|
|
switch err {
|
|
|
|
case nil:
|
|
|
|
case storage.ErrOutOfOrderSample, storage.ErrDuplicateSampleForTimestamp:
|
|
|
|
// Do not count these in logging, as this is expected if series
|
|
|
|
// is exposed from a different group.
|
|
|
|
continue
|
|
|
|
default:
|
|
|
|
log.With("sample", metric).With("err", err).Warn("adding stale sample failed")
|
|
|
|
if err := app.Rollback(); err != nil {
|
|
|
|
log.With("err", err).Warn("rule stale sample rollback failed")
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if err := app.Commit(); err != nil {
|
|
|
|
log.With("err", err).Warn("rule stale sample appending failed")
|
|
|
|
}
|
|
|
|
g.seriesInPreviousEval = seriesReturned
|
2015-12-14 16:40:40 +00:00
|
|
|
}
|
|
|
|
|
2015-12-17 10:46:10 +00:00
|
|
|
// sendAlerts sends alert notifications for the given rule.
|
2016-12-29 16:31:14 +00:00
|
|
|
func (g *Group) sendAlerts(rule *AlertingRule) error {
|
2016-12-24 23:37:46 +00:00
|
|
|
var alerts []*notifier.Alert
|
2015-12-15 18:46:03 +00:00
|
|
|
|
2015-12-17 10:46:10 +00:00
|
|
|
for _, alert := range rule.currentAlerts() {
|
2015-12-14 16:40:40 +00:00
|
|
|
// Only send actually firing alerts.
|
2015-12-15 18:46:03 +00:00
|
|
|
if alert.State == StatePending {
|
2013-07-30 15:18:07 +00:00
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
2016-12-24 23:37:46 +00:00
|
|
|
a := ¬ifier.Alert{
|
2016-12-29 16:31:14 +00:00
|
|
|
StartsAt: alert.ActiveAt.Add(rule.holdDuration),
|
2016-07-12 16:11:31 +00:00
|
|
|
Labels: alert.Labels,
|
|
|
|
Annotations: alert.Annotations,
|
2015-12-14 16:40:40 +00:00
|
|
|
GeneratorURL: g.opts.ExternalURL.String() + strutil.GraphLinkForExpression(rule.vector.String()),
|
2015-12-15 18:46:03 +00:00
|
|
|
}
|
2016-12-29 16:31:14 +00:00
|
|
|
if !alert.ResolvedAt.IsZero() {
|
|
|
|
a.EndsAt = alert.ResolvedAt
|
2015-12-15 18:46:03 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
alerts = append(alerts, a)
|
2013-07-30 15:18:07 +00:00
|
|
|
}
|
2013-05-23 19:29:27 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
if len(alerts) > 0 {
|
2016-03-01 11:37:22 +00:00
|
|
|
g.opts.Notifier.Send(alerts...)
|
2015-12-14 16:40:40 +00:00
|
|
|
}
|
2013-12-10 15:54:35 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
return nil
|
|
|
|
}
|
2013-12-10 15:54:35 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
// The Manager manages recording and alerting rules.
|
|
|
|
type Manager struct {
|
|
|
|
opts *ManagerOptions
|
|
|
|
groups map[string]*Group
|
|
|
|
mtx sync.RWMutex
|
2016-01-08 16:51:22 +00:00
|
|
|
block chan struct{}
|
2015-12-14 16:40:40 +00:00
|
|
|
}
|
2013-07-30 15:18:07 +00:00
|
|
|
|
2017-01-13 13:48:01 +00:00
|
|
|
type Appendable interface {
|
|
|
|
Appender() (storage.Appender, error)
|
|
|
|
}
|
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
// ManagerOptions bundles options for the Manager.
|
|
|
|
type ManagerOptions struct {
|
2017-01-13 13:48:01 +00:00
|
|
|
ExternalURL *url.URL
|
|
|
|
QueryEngine *promql.Engine
|
|
|
|
Context context.Context
|
|
|
|
Notifier *notifier.Notifier
|
|
|
|
Appendable Appendable
|
2015-12-14 16:40:40 +00:00
|
|
|
}
|
2015-03-15 02:36:15 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
// NewManager returns an implementation of Manager, ready to be started
|
|
|
|
// by calling the Run method.
|
|
|
|
func NewManager(o *ManagerOptions) *Manager {
|
|
|
|
manager := &Manager{
|
|
|
|
groups: map[string]*Group{},
|
|
|
|
opts: o,
|
2016-01-08 16:51:22 +00:00
|
|
|
block: make(chan struct{}),
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
2015-12-14 16:40:40 +00:00
|
|
|
return manager
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
|
|
|
|
2016-01-08 16:51:22 +00:00
|
|
|
// Run starts processing of the rule manager.
|
|
|
|
func (m *Manager) Run() {
|
|
|
|
close(m.block)
|
|
|
|
}
|
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
// Stop the rule manager's rule evaluation cycles.
|
|
|
|
func (m *Manager) Stop() {
|
rules/manager.go: Fix race between reload and stop
On one relatively large Prometheus instance (1.7M series), I noticed
that upgrades were frequently resulting in Prometheus undergoing crash
recovery on start-up.
On closer examination, I found that Prometheus was panicking on
shutdown.
It seems that our configuration management (or misconfiguration thereof)
is reloading Prometheus then immediately restarting it, which I suspect
is causing this race:
Sep 21 15:12:42 host systemd[1]: Reloading prometheus monitoring system.
Sep 21 15:12:42 host prometheus[18734]: time="2016-09-21T15:12:42Z" level=info msg="Loading configuration file /etc/prometheus/config.yaml" source="main.go:221"
Sep 21 15:12:42 host systemd[1]: Reloaded prometheus monitoring system.
Sep 21 15:12:44 host systemd[1]: Stopping prometheus monitoring system...
Sep 21 15:12:44 host prometheus[18734]: time="2016-09-21T15:12:44Z" level=warning msg="Received SIGTERM, exiting gracefully..." source="main.go:203"
Sep 21 15:12:44 host prometheus[18734]: time="2016-09-21T15:12:44Z" level=info msg="See you next time!" source="main.go:210"
Sep 21 15:12:44 host prometheus[18734]: time="2016-09-21T15:12:44Z" level=info msg="Stopping target manager..." source="targetmanager.go:90"
Sep 21 15:12:52 host prometheus[18734]: time="2016-09-21T15:12:52Z" level=info msg="Checkpointing in-memory metrics and chunks..." source="persistence.go:548"
Sep 21 15:12:56 host prometheus[18734]: time="2016-09-21T15:12:56Z" level=warning msg="Error on ingesting out-of-order samples" numDropped=1 source="scrape.go:467"
Sep 21 15:12:56 host prometheus[18734]: time="2016-09-21T15:12:56Z" level=error msg="Error adding file watch for \"/etc/prometheus/targets\": no such file or directory" source="file.go:84"
Sep 21 15:12:56 host prometheus[18734]: time="2016-09-21T15:12:56Z" level=error msg="Error adding file watch for \"/etc/prometheus/targets\": no such file or directory" source="file.go:84"
Sep 21 15:13:01 host prometheus[18734]: time="2016-09-21T15:13:01Z" level=info msg="Stopping rule manager..." source="manager.go:366"
Sep 21 15:13:01 host prometheus[18734]: time="2016-09-21T15:13:01Z" level=info msg="Rule manager stopped." source="manager.go:372"
Sep 21 15:13:01 host prometheus[18734]: time="2016-09-21T15:13:01Z" level=info msg="Stopping notification handler..." source="notifier.go:325"
Sep 21 15:13:01 host prometheus[18734]: time="2016-09-21T15:13:01Z" level=info msg="Stopping local storage..." source="storage.go:381"
Sep 21 15:13:01 host prometheus[18734]: time="2016-09-21T15:13:01Z" level=info msg="Stopping maintenance loop..." source="storage.go:383"
Sep 21 15:13:01 host prometheus[18734]: panic: close of closed channel
Sep 21 15:13:01 host prometheus[18734]: goroutine 7686074 [running]:
Sep 21 15:13:01 host prometheus[18734]: panic(0xba57a0, 0xc60c42b500)
Sep 21 15:13:01 host prometheus[18734]: /usr/local/go/src/runtime/panic.go:500 +0x1a1
Sep 21 15:13:01 host prometheus[18734]: github.com/prometheus/prometheus/rules.(*Manager).ApplyConfig.func1(0xc6645a9901, 0xc420271ef0, 0xc420338ed0, 0xc60c42b4f0, 0xc6645a9900)
Sep 21 15:13:01 host prometheus[18734]: /home/build/packages/prometheus/tmp/build/gopath/src/github.com/prometheus/prometheus/rules/manager.go:412 +0x3c
Sep 21 15:13:01 host prometheus[18734]: created by github.com/prometheus/prometheus/rules.(*Manager).ApplyConfig
Sep 21 15:13:01 host prometheus[18734]: /home/build/packages/prometheus/tmp/build/gopath/src/github.com/prometheus/prometheus/rules/manager.go:423 +0x56b
Sep 21 15:13:03 host systemd[1]: prometheus.service: main process exited, code=exited, status=2/INVALIDARGUMENT
2016-09-21 21:03:02 +00:00
|
|
|
m.mtx.Lock()
|
|
|
|
defer m.mtx.Unlock()
|
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
log.Info("Stopping rule manager...")
|
2015-06-30 09:51:05 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
for _, eg := range m.groups {
|
|
|
|
eg.stop()
|
2015-06-30 09:51:05 +00:00
|
|
|
}
|
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
log.Info("Rule manager stopped.")
|
2015-06-30 09:51:05 +00:00
|
|
|
}
|
|
|
|
|
2015-05-12 14:52:56 +00:00
|
|
|
// ApplyConfig updates the rule manager's state as the config requires. If
|
2016-07-11 14:24:54 +00:00
|
|
|
// loading the new rules failed the old rule set is restored.
|
|
|
|
func (m *Manager) ApplyConfig(conf *config.Config) error {
|
2015-12-14 16:40:40 +00:00
|
|
|
m.mtx.Lock()
|
|
|
|
defer m.mtx.Unlock()
|
2015-05-12 14:52:56 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
// Get all rule files and load the groups they define.
|
2015-05-28 05:36:21 +00:00
|
|
|
var files []string
|
|
|
|
for _, pat := range conf.RuleFiles {
|
|
|
|
fs, err := filepath.Glob(pat)
|
|
|
|
if err != nil {
|
|
|
|
// The only error can be a bad pattern.
|
2016-07-11 14:24:54 +00:00
|
|
|
return fmt.Errorf("error retrieving rule files for %s: %s", pat, err)
|
2015-05-28 05:36:21 +00:00
|
|
|
}
|
|
|
|
files = append(files, fs...)
|
|
|
|
}
|
2015-12-14 16:40:40 +00:00
|
|
|
|
2016-11-18 16:32:30 +00:00
|
|
|
// To be replaced with a configurable per-group interval.
|
|
|
|
groups, err := m.loadGroups(time.Duration(conf.GlobalConfig.EvaluationInterval), files...)
|
2015-12-14 16:40:40 +00:00
|
|
|
if err != nil {
|
2016-07-11 14:24:54 +00:00
|
|
|
return fmt.Errorf("error loading rules, previous rule set restored: %s", err)
|
2015-05-12 14:52:56 +00:00
|
|
|
}
|
2015-06-23 10:07:53 +00:00
|
|
|
|
2015-12-14 16:40:40 +00:00
|
|
|
var wg sync.WaitGroup
|
|
|
|
|
|
|
|
for _, newg := range groups {
|
|
|
|
wg.Add(1)
|
|
|
|
|
|
|
|
// If there is an old group with the same identifier, stop it and wait for
|
2017-05-18 15:26:36 +00:00
|
|
|
// it to finish the current iteration. Then copy it into the new group.
|
2015-12-14 16:40:40 +00:00
|
|
|
oldg, ok := m.groups[newg.name]
|
|
|
|
delete(m.groups, newg.name)
|
|
|
|
|
|
|
|
go func(newg *Group) {
|
|
|
|
if ok {
|
|
|
|
oldg.stop()
|
|
|
|
newg.copyState(oldg)
|
|
|
|
}
|
2016-01-08 16:51:22 +00:00
|
|
|
go func() {
|
|
|
|
// Wait with starting evaluation until the rule manager
|
|
|
|
// is told to run. This is necessary to avoid running
|
|
|
|
// queries against a bootstrapping storage.
|
|
|
|
<-m.block
|
|
|
|
newg.run()
|
|
|
|
}()
|
2015-12-14 16:40:40 +00:00
|
|
|
wg.Done()
|
|
|
|
}(newg)
|
|
|
|
}
|
|
|
|
|
|
|
|
// Stop remaining old groups.
|
|
|
|
for _, oldg := range m.groups {
|
|
|
|
oldg.stop()
|
|
|
|
}
|
|
|
|
|
|
|
|
wg.Wait()
|
|
|
|
m.groups = groups
|
|
|
|
|
2016-07-11 14:24:54 +00:00
|
|
|
return nil
|
2015-05-12 14:52:56 +00:00
|
|
|
}
|
|
|
|
|
2015-12-17 10:46:10 +00:00
|
|
|
// loadGroups reads groups from a list of files.
|
|
|
|
// As there's currently no group syntax a single group named "default" containing
|
|
|
|
// all rules will be returned.
|
2016-11-18 16:32:30 +00:00
|
|
|
func (m *Manager) loadGroups(interval time.Duration, filenames ...string) (map[string]*Group, error) {
|
2016-11-18 16:25:58 +00:00
|
|
|
rules := []Rule{}
|
2015-04-29 09:08:56 +00:00
|
|
|
for _, fn := range filenames {
|
|
|
|
content, err := ioutil.ReadFile(fn)
|
|
|
|
if err != nil {
|
2015-12-14 16:40:40 +00:00
|
|
|
return nil, err
|
2015-04-29 09:08:56 +00:00
|
|
|
}
|
|
|
|
stmts, err := promql.ParseStmts(string(content))
|
|
|
|
if err != nil {
|
2015-12-14 16:40:40 +00:00
|
|
|
return nil, fmt.Errorf("error parsing %s: %s", fn, err)
|
2015-04-29 09:08:56 +00:00
|
|
|
}
|
2015-07-03 12:48:22 +00:00
|
|
|
|
2015-04-29 09:08:56 +00:00
|
|
|
for _, stmt := range stmts {
|
2015-12-14 16:40:40 +00:00
|
|
|
var rule Rule
|
|
|
|
|
2015-04-29 09:08:56 +00:00
|
|
|
switch r := stmt.(type) {
|
|
|
|
case *promql.AlertStmt:
|
2015-12-14 16:40:40 +00:00
|
|
|
rule = NewAlertingRule(r.Name, r.Expr, r.Duration, r.Labels, r.Annotations)
|
|
|
|
|
2015-04-29 09:08:56 +00:00
|
|
|
case *promql.RecordStmt:
|
2015-12-14 16:40:40 +00:00
|
|
|
rule = NewRecordingRule(r.Name, r.Expr, r.Labels)
|
|
|
|
|
2015-04-29 09:08:56 +00:00
|
|
|
default:
|
|
|
|
panic("retrieval.Manager.LoadRuleFiles: unknown statement type")
|
|
|
|
}
|
2016-11-18 16:25:58 +00:00
|
|
|
rules = append(rules, rule)
|
2015-04-29 09:08:56 +00:00
|
|
|
}
|
|
|
|
}
|
2015-12-14 16:40:40 +00:00
|
|
|
|
2016-11-18 16:25:58 +00:00
|
|
|
// Currently there is no group syntax implemented. Thus all rules
|
|
|
|
// are read into a single default group.
|
2016-11-18 16:32:30 +00:00
|
|
|
g := NewGroup("default", interval, rules, m.opts)
|
2016-11-18 16:25:58 +00:00
|
|
|
groups := map[string]*Group{g.name: g}
|
2015-12-14 16:40:40 +00:00
|
|
|
return groups, nil
|
2013-01-07 22:24:26 +00:00
|
|
|
}
|
2013-06-11 09:00:55 +00:00
|
|
|
|
2015-04-29 09:08:56 +00:00
|
|
|
// Rules returns the list of the manager's rules.
|
2015-04-29 08:26:49 +00:00
|
|
|
func (m *Manager) Rules() []Rule {
|
2015-12-14 16:40:40 +00:00
|
|
|
m.mtx.RLock()
|
|
|
|
defer m.mtx.RUnlock()
|
|
|
|
|
|
|
|
var rules []Rule
|
|
|
|
for _, g := range m.groups {
|
|
|
|
rules = append(rules, g.rules...)
|
|
|
|
}
|
2013-06-11 09:00:55 +00:00
|
|
|
|
|
|
|
return rules
|
|
|
|
}
|
2013-06-13 14:10:05 +00:00
|
|
|
|
2015-04-29 09:08:56 +00:00
|
|
|
// AlertingRules returns the list of the manager's alerting rules.
|
2015-04-29 08:26:49 +00:00
|
|
|
func (m *Manager) AlertingRules() []*AlertingRule {
|
2015-12-14 16:40:40 +00:00
|
|
|
m.mtx.RLock()
|
|
|
|
defer m.mtx.RUnlock()
|
2013-06-13 14:10:05 +00:00
|
|
|
|
2015-03-30 17:43:19 +00:00
|
|
|
alerts := []*AlertingRule{}
|
2015-12-14 16:40:40 +00:00
|
|
|
for _, rule := range m.Rules() {
|
2015-03-30 17:43:19 +00:00
|
|
|
if alertingRule, ok := rule.(*AlertingRule); ok {
|
2013-06-13 14:10:05 +00:00
|
|
|
alerts = append(alerts, alertingRule)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return alerts
|
|
|
|
}
|