prometheus/rules/ast/query_analyzer.go

161 lines
6.0 KiB
Go
Raw Normal View History

// Copyright 2013 Prometheus Team
// 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.
package ast
import (
"time"
"github.com/golang/glog"
clientmodel "github.com/prometheus/client_golang/model"
"github.com/prometheus/prometheus/stats"
"github.com/prometheus/prometheus/storage/metric"
)
// FullRangeMap maps the fingerprint of a full range to the duration
// of the matrix selector it resulted from.
type FullRangeMap map[clientmodel.Fingerprint]time.Duration
// IntervalRangeMap is a set of fingerprints of interval ranges.
type IntervalRangeMap map[clientmodel.Fingerprint]bool
// A QueryAnalyzer recursively traverses the AST to look for any nodes
// which will need data from the datastore. Instantiate with
// NewQueryAnalyzer.
type QueryAnalyzer struct {
// Values collected by query analysis.
//
// Full ranges always implicitly span a time range of:
// - start: query interval start - duration
// - end: query interval end
//
// This is because full ranges can only result from matrix selectors (like
// "foo[5m]"), which have said time-spanning behavior during a ranged query.
FullRanges FullRangeMap
// Interval ranges always implicitly span the whole query range.
IntervalRanges IntervalRangeMap
2013-05-07 11:15:10 +00:00
// The underlying storage to which the query will be applied. Needed for
// extracting timeseries fingerprint information during query analysis.
storage *metric.TieredStorage
}
// NewQueryAnalyzer returns a pointer to a newly instantiated
// QueryAnalyzer. The storage is needed to extract timeseries
// fingerprint information during query analysis.
2013-05-07 11:15:10 +00:00
func NewQueryAnalyzer(storage *metric.TieredStorage) *QueryAnalyzer {
return &QueryAnalyzer{
FullRanges: FullRangeMap{},
IntervalRanges: IntervalRangeMap{},
2013-05-07 11:15:10 +00:00
storage: storage,
}
}
// Visit implements the Visitor interface.
func (analyzer *QueryAnalyzer) Visit(node Node) {
switch n := node.(type) {
case *VectorSelector:
2013-05-07 11:15:10 +00:00
fingerprints, err := analyzer.storage.GetFingerprintsForLabelSet(n.labels)
if err != nil {
glog.Errorf("Error getting fingerprints for labelset %v: %v", n.labels, err)
return
}
n.fingerprints = fingerprints
for _, fingerprint := range fingerprints {
// Only add the fingerprint to IntervalRanges if not yet present in FullRanges.
// Full ranges always contain more points and span more time than interval ranges.
if _, alreadyInFullRanges := analyzer.FullRanges[*fingerprint]; !alreadyInFullRanges {
analyzer.IntervalRanges[*fingerprint] = true
}
}
case *MatrixSelector:
2013-05-07 11:15:10 +00:00
fingerprints, err := analyzer.storage.GetFingerprintsForLabelSet(n.labels)
if err != nil {
glog.Errorf("Error getting fingerprints for labelset %v: %v", n.labels, err)
return
}
n.fingerprints = fingerprints
for _, fingerprint := range fingerprints {
if analyzer.FullRanges[*fingerprint] < n.interval {
analyzer.FullRanges[*fingerprint] = n.interval
// Delete the fingerprint from IntervalRanges. Full ranges always contain
// more points and span more time than interval ranges, so we don't need
// an interval range for the same fingerprint, should we have one.
delete(analyzer.IntervalRanges, *fingerprint)
}
}
}
}
// AnalyzeQueries walks the AST, starting at node, calling Visit on
// each node to collect fingerprints.
func (analyzer *QueryAnalyzer) AnalyzeQueries(node Node) {
Walk(analyzer, node)
}
Use custom timestamp type for sample timestamps and related code. So far we've been using Go's native time.Time for anything related to sample timestamps. Since the range of time.Time is much bigger than what we need, this has created two problems: - there could be time.Time values which were out of the range/precision of the time type that we persist to disk, therefore causing incorrectly ordered keys. One bug caused by this was: https://github.com/prometheus/prometheus/issues/367 It would be good to use a timestamp type that's more closely aligned with what the underlying storage supports. - sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit Unix timestamp (possibly even a 32-bit one). Since we store samples in large numbers, this seriously affects memory usage. Furthermore, copying/working with the data will be faster if it's smaller. *MEMORY USAGE RESULTS* Initial memory usage comparisons for a running Prometheus with 1 timeseries and 100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my tests, this advantage for some reason decreased a bit the more samples the timeseries had (to 5-7% for millions of samples). This I can't fully explain, but perhaps garbage collection issues were involved. *WHEN TO USE THE NEW TIMESTAMP TYPE* The new clientmodel.Timestamp type should be used whenever time calculations are either directly or indirectly related to sample timestamps. For example: - the timestamp of a sample itself - all kinds of watermarks - anything that may become or is compared to a sample timestamp (like the timestamp passed into Target.Scrape()). When to still use time.Time: - for measuring durations/times not related to sample timestamps, like duration telemetry exporting, timers that indicate how frequently to execute some action, etc. *NOTE ON OPERATOR OPTIMIZATION TESTS* We don't use operator optimization code anymore, but it still lives in the code as dead code. It still has tests, but I couldn't get all of them to pass with the new timestamp format. I commented out the failing cases for now, but we should probably remove the dead code soon. I just didn't want to do that in the same change as this. Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
func viewAdapterForInstantQuery(node Node, timestamp clientmodel.Timestamp, storage *metric.TieredStorage, queryStats *stats.TimerGroup) (*viewAdapter, error) {
analyzeTimer := queryStats.GetTimer(stats.QueryAnalysisTime).Start()
2013-05-07 11:15:10 +00:00
analyzer := NewQueryAnalyzer(storage)
analyzer.AnalyzeQueries(node)
analyzeTimer.Stop()
requestBuildTimer := queryStats.GetTimer(stats.ViewRequestBuildTime).Start()
viewBuilder := metric.NewViewRequestBuilder()
for fingerprint, rangeDuration := range analyzer.FullRanges {
viewBuilder.GetMetricRange(&fingerprint, timestamp.Add(-rangeDuration), timestamp)
}
for fingerprint := range analyzer.IntervalRanges {
viewBuilder.GetMetricAtTime(&fingerprint, timestamp)
}
requestBuildTimer.Stop()
buildTimer := queryStats.GetTimer(stats.InnerViewBuildingTime).Start()
// BUG(julius): Clear Law of Demeter violation.
view, err := analyzer.storage.MakeView(viewBuilder, 60*time.Second, queryStats)
buildTimer.Stop()
if err != nil {
return nil, err
}
return NewViewAdapter(view, storage, queryStats), nil
}
Use custom timestamp type for sample timestamps and related code. So far we've been using Go's native time.Time for anything related to sample timestamps. Since the range of time.Time is much bigger than what we need, this has created two problems: - there could be time.Time values which were out of the range/precision of the time type that we persist to disk, therefore causing incorrectly ordered keys. One bug caused by this was: https://github.com/prometheus/prometheus/issues/367 It would be good to use a timestamp type that's more closely aligned with what the underlying storage supports. - sizeof(time.Time) is 192, while Prometheus should be ok with a single 64-bit Unix timestamp (possibly even a 32-bit one). Since we store samples in large numbers, this seriously affects memory usage. Furthermore, copying/working with the data will be faster if it's smaller. *MEMORY USAGE RESULTS* Initial memory usage comparisons for a running Prometheus with 1 timeseries and 100,000 samples show roughly a 13% decrease in total (VIRT) memory usage. In my tests, this advantage for some reason decreased a bit the more samples the timeseries had (to 5-7% for millions of samples). This I can't fully explain, but perhaps garbage collection issues were involved. *WHEN TO USE THE NEW TIMESTAMP TYPE* The new clientmodel.Timestamp type should be used whenever time calculations are either directly or indirectly related to sample timestamps. For example: - the timestamp of a sample itself - all kinds of watermarks - anything that may become or is compared to a sample timestamp (like the timestamp passed into Target.Scrape()). When to still use time.Time: - for measuring durations/times not related to sample timestamps, like duration telemetry exporting, timers that indicate how frequently to execute some action, etc. *NOTE ON OPERATOR OPTIMIZATION TESTS* We don't use operator optimization code anymore, but it still lives in the code as dead code. It still has tests, but I couldn't get all of them to pass with the new timestamp format. I commented out the failing cases for now, but we should probably remove the dead code soon. I just didn't want to do that in the same change as this. Change-Id: I821787414b0debe85c9fffaeb57abd453727af0f
2013-10-28 13:35:02 +00:00
func viewAdapterForRangeQuery(node Node, start clientmodel.Timestamp, end clientmodel.Timestamp, interval time.Duration, storage *metric.TieredStorage, queryStats *stats.TimerGroup) (*viewAdapter, error) {
analyzeTimer := queryStats.GetTimer(stats.QueryAnalysisTime).Start()
2013-05-07 11:15:10 +00:00
analyzer := NewQueryAnalyzer(storage)
analyzer.AnalyzeQueries(node)
analyzeTimer.Stop()
requestBuildTimer := queryStats.GetTimer(stats.ViewRequestBuildTime).Start()
viewBuilder := metric.NewViewRequestBuilder()
for fingerprint, rangeDuration := range analyzer.FullRanges {
if interval < rangeDuration {
viewBuilder.GetMetricRange(&fingerprint, start.Add(-rangeDuration), end)
} else {
viewBuilder.GetMetricRangeAtInterval(&fingerprint, start.Add(-rangeDuration), end, interval, rangeDuration)
}
}
for fingerprint := range analyzer.IntervalRanges {
viewBuilder.GetMetricAtInterval(&fingerprint, start, end, interval)
}
requestBuildTimer.Stop()
buildTimer := queryStats.GetTimer(stats.InnerViewBuildingTime).Start()
view, err := analyzer.storage.MakeView(viewBuilder, time.Duration(60)*time.Second, queryStats)
buildTimer.Stop()
if err != nil {
return nil, err
}
return NewViewAdapter(view, storage, queryStats), nil
}