github-mirrors/VictoriaMetrics
1
0
Fork 0
mirror of https://github.com/VictoriaMetrics/VictoriaMetrics.git synced 2024-11-21 14:44:00 +00:00
Code Issues Projects Releases Packages Wiki Activity

lib/storage: increase match cost for negative tag filters, since they need to scan all the label pairs

Browse source
This commit is contained in:
Aliaksandr Valialkin 2021-02-15 16:24:08 +02:00
parent 4e39bf148c
commit ecf132933e
2 changed files with 15 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
lib/storage/index_db.go
View file

@ -2804,7 +2804,7 @@ func (is *indexSearch) getMetricIDsForDateAndFilters(date uint64, tfs *TagFilter
if a.seconds != b.seconds { if a.seconds != b.seconds {
return a.seconds < b.seconds return a.seconds < b.seconds
} }
return a.tf.matchCost < b.tf.matchCost return a.tf.Less(b.tf)
}) })
// Populate metricIDs for the first non-negative filter. // Populate metricIDs for the first non-negative filter.

14
lib/storage/tag_filters.go
View file

@ -240,6 +240,9 @@ type tagFilter struct {
func (tf *tagFilter) Less(other *tagFilter) bool { func (tf *tagFilter) Less(other *tagFilter) bool {
// Move regexp and negative filters to the end, since they require scanning // Move regexp and negative filters to the end, since they require scanning
// all the entries for the given label. // all the entries for the given label.
if tf.matchCost != other.matchCost {
return tf.matchCost < other.matchCost
}
if tf.isRegexp != other.isRegexp { if tf.isRegexp != other.isRegexp {
return !tf.isRegexp return !tf.isRegexp
} }
@ -312,6 +315,9 @@ func (tf *tagFilter) InitFromGraphiteQuery(commonPrefix, query []byte, paths []s
tf.prefix = marshalTagValueNoTrailingTagSeparator(tf.prefix, []byte(prefix)) tf.prefix = marshalTagValueNoTrailingTagSeparator(tf.prefix, []byte(prefix))
tf.orSuffixes = append(tf.orSuffixes[:0], orSuffixes...) tf.orSuffixes = append(tf.orSuffixes[:0], orSuffixes...)
tf.reSuffixMatch, tf.matchCost = newMatchFuncForOrSuffixes(orSuffixes) tf.reSuffixMatch, tf.matchCost = newMatchFuncForOrSuffixes(orSuffixes)
if isNegative {
tf.matchCost *= negativeMatchCostMultiplier
}
} }
func getCommonPrefix(ss []string) (string, []string) { func getCommonPrefix(ss []string) (string, []string) {
@ -379,6 +385,9 @@ func (tf *tagFilter) Init(commonPrefix, key, value []byte, isNegative, isRegexp
tf.orSuffixes = append(tf.orSuffixes[:0], "") tf.orSuffixes = append(tf.orSuffixes[:0], "")
tf.isEmptyMatch = len(prefix) == 0 tf.isEmptyMatch = len(prefix) == 0
tf.matchCost = fullMatchCost tf.matchCost = fullMatchCost
if isNegative {
tf.matchCost *= negativeMatchCostMultiplier
}
return nil return nil
} }
rcv, err := getRegexpFromCache(expr) rcv, err := getRegexpFromCache(expr)
@ -388,6 +397,9 @@ func (tf *tagFilter) Init(commonPrefix, key, value []byte, isNegative, isRegexp
tf.orSuffixes = append(tf.orSuffixes[:0], rcv.orValues...) tf.orSuffixes = append(tf.orSuffixes[:0], rcv.orValues...)
tf.reSuffixMatch = rcv.reMatch tf.reSuffixMatch = rcv.reMatch
tf.matchCost = rcv.reCost tf.matchCost = rcv.reCost
if isNegative {
tf.matchCost *= negativeMatchCostMultiplier
}
tf.isEmptyMatch = len(prefix) == 0 && tf.reSuffixMatch(nil) tf.isEmptyMatch = len(prefix) == 0 && tf.reSuffixMatch(nil)
if !tf.isNegative && len(key) == 0 && strings.IndexByte(rcv.literalSuffix, '.') >= 0 { if !tf.isNegative && len(key) == 0 && strings.IndexByte(rcv.literalSuffix, '.') >= 0 {
// Reverse suffix is needed only for non-negative regexp filters on __name__ that contains dots. // Reverse suffix is needed only for non-negative regexp filters on __name__ that contains dots.
@ -559,6 +571,8 @@ const (
reMatchCost = 100 reMatchCost = 100
) )
const negativeMatchCostMultiplier = 1000
func getOptimizedReMatchFuncExt(reMatch func(b []byte) bool, sre *syntax.Regexp) (func(b []byte) bool, string, uint64) { func getOptimizedReMatchFuncExt(reMatch func(b []byte) bool, sre *syntax.Regexp) (func(b []byte) bool, string, uint64) {
if isDotStar(sre) { if isDotStar(sre) {
// '.*' // '.*'