Skip to content

Fix Cache to use variable labels #780

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jul 24, 2024
Merged

Fix Cache to use variable labels #780

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d99bf47
Fix Cache to use variable labels
lucacome Jul 17, 2024
28ab7ca
Merge branch 'main' into fix/cache
lucacome Jul 23, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
121 changes: 70 additions & 51 deletions collector/nginx_plus.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -223,6 +223,12 @@
return c.streamUpstreamServerPeerLabels[peer]
}

func (c *NginxPlusCollector) getCacheZoneLabelValues(cacheName string) []string {
c.variableLabelsMutex.RLock()
defer c.variableLabelsMutex.RUnlock()
return c.cacheZoneLabels[cacheName]

Check warning on line 229 in collector/nginx_plus.go

View check run for this annotation

Codecov / codecov/patch

collector/nginx_plus.go#L226-L229 

Added lines #L226 - L229 were not covered by tests
}

// VariableLabelNames holds all the variable label names for the different metrics.
type VariableLabelNames struct {
UpstreamServerVariableLabelNames []string
Expand All @@ -231,12 +237,12 @@
StreamUpstreamServerPeerVariableLabelNames []string
StreamServerZoneVariableLabelNames []string
StreamUpstreamServerVariableLabelNames []string
CacheZoneLabelNames []string
CacheZoneVariableLabelNames []string
}

// NewVariableLabelNames NewVariableLabels creates a new struct for VariableNames for the collector.
func NewVariableLabelNames(upstreamServerVariableLabelNames []string, serverZoneVariableLabelNames []string, upstreamServerPeerVariableLabelNames []string,
streamUpstreamServerVariableLabelNames []string, streamServerZoneLabels []string, streamUpstreamServerPeerVariableLabelNames []string, cacheZoneLabelNames []string,
streamUpstreamServerVariableLabelNames []string, streamServerZoneLabels []string, streamUpstreamServerPeerVariableLabelNames []string, cacheZoneVariableLabelNames []string,
) VariableLabelNames {
return VariableLabelNames{
UpstreamServerVariableLabelNames: upstreamServerVariableLabelNames,
Expand All @@ -245,7 +251,7 @@
StreamUpstreamServerVariableLabelNames: streamUpstreamServerVariableLabelNames,
StreamServerZoneVariableLabelNames: streamServerZoneLabels,
StreamUpstreamServerPeerVariableLabelNames: streamUpstreamServerPeerVariableLabelNames,
CacheZoneLabelNames: cacheZoneLabelNames,
CacheZoneVariableLabelNames: cacheZoneVariableLabelNames,

Check warning on line 254 in collector/nginx_plus.go

View check run for this annotation

Codecov / codecov/patch

collector/nginx_plus.go#L254 

Added line #L254 was not covered by tests
}
}

Expand All @@ -264,6 +270,7 @@
upstreamServerPeerLabels: make(map[string][]string),
streamUpstreamServerPeerLabels: make(map[string][]string),
streamUpstreamServerLabels: make(map[string][]string),
cacheZoneLabels: make(map[string][]string),

Check warning on line 273 in collector/nginx_plus.go

View check run for this annotation

Codecov / codecov/patch

collector/nginx_plus.go#L273 

Added line #L273 was not covered by tests
nginxClient: nginxClient,
logger: logger,
totalMetrics: map[string]*prometheus.Desc{
Expand Down Expand Up @@ -526,27 +533,27 @@
},
upMetric: newUpMetric(namespace, constLabels),
cacheZoneMetrics: map[string]*prometheus.Desc{
"size": newCacheZoneMetric(namespace, "size", "Total size of the cache", constLabels),
"max_size": newCacheZoneMetric(namespace, "max_size", "Maximum size of the cache", constLabels),
"cold": newCacheZoneMetric(namespace, "cold", "Is the cache considered cold", constLabels),
"hit_responses": newCacheZoneMetric(namespace, "hit_responses", "Total number of cache hits", constLabels),
"hit_bytes": newCacheZoneMetric(namespace, "hit_bytes", "Total number of bytes returned from cache", constLabels),
"stale_responses": newCacheZoneMetric(namespace, "stale_responses", "Total number of stale cache hits", constLabels),
"stale_bytes": newCacheZoneMetric(namespace, "stale_bytes", "Total number of bytes returned from stale cache", constLabels),
"updating_responses": newCacheZoneMetric(namespace, "updating_responses", "Total number of cache hits while cache is updating", constLabels),
"updating_bytes": newCacheZoneMetric(namespace, "updating_bytes", "Total number of bytes returned from cache while cache is updating", constLabels),
"revalidated_responses": newCacheZoneMetric(namespace, "revalidated_responses", "Total number of cache revalidations", constLabels),
"revalidated_bytes": newCacheZoneMetric(namespace, "revalidated_bytes", "Total number of bytes returned from cache revalidations", constLabels),
"miss_responses": newCacheZoneMetric(namespace, "miss_responses", "Total number of cache misses", constLabels),
"miss_bytes": newCacheZoneMetric(namespace, "miss_bytes", "Total number of bytes returned from cache misses", constLabels),
"expired_responses": newCacheZoneMetric(namespace, "expired_responses", "Total number of cache hits with expired TTL", constLabels),
"expired_bytes": newCacheZoneMetric(namespace, "expired_bytes", "Total number of bytes returned from cache hits with expired TTL", constLabels),
"expired_responses_written": newCacheZoneMetric(namespace, "expired_responses_written", "Total number of cache hits with expired TTL written to cache", constLabels),
"expired_bytes_written": newCacheZoneMetric(namespace, "expired_bytes_written", "Total number of bytes written to cache from cache hits with expired TTL", constLabels),
"bypass_responses": newCacheZoneMetric(namespace, "bypass_responses", "Total number of cache bypasses", constLabels),
"bypass_bytes": newCacheZoneMetric(namespace, "bypass_bytes", "Total number of bytes returned from cache bypasses", constLabels),
"bypass_responses_written": newCacheZoneMetric(namespace, "bypass_responses_written", "Total number of cache bypasses written to cache", constLabels),
"bypass_bytes_written": newCacheZoneMetric(namespace, "bypass_bytes_written", "Total number of bytes written to cache from cache bypasses", constLabels),
"size": newCacheZoneMetric(namespace, "size", "Total size of the cache", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"max_size": newCacheZoneMetric(namespace, "max_size", "Maximum size of the cache", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"cold": newCacheZoneMetric(namespace, "cold", "Is the cache considered cold", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"hit_responses": newCacheZoneMetric(namespace, "hit_responses", "Total number of cache hits", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"hit_bytes": newCacheZoneMetric(namespace, "hit_bytes", "Total number of bytes returned from cache", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"stale_responses": newCacheZoneMetric(namespace, "stale_responses", "Total number of stale cache hits", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"stale_bytes": newCacheZoneMetric(namespace, "stale_bytes", "Total number of bytes returned from stale cache", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"updating_responses": newCacheZoneMetric(namespace, "updating_responses", "Total number of cache hits while cache is updating", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"updating_bytes": newCacheZoneMetric(namespace, "updating_bytes", "Total number of bytes returned from cache while cache is updating", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"revalidated_responses": newCacheZoneMetric(namespace, "revalidated_responses", "Total number of cache revalidations", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"revalidated_bytes": newCacheZoneMetric(namespace, "revalidated_bytes", "Total number of bytes returned from cache revalidations", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"miss_responses": newCacheZoneMetric(namespace, "miss_responses", "Total number of cache misses", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"miss_bytes": newCacheZoneMetric(namespace, "miss_bytes", "Total number of bytes returned from cache misses", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"expired_responses": newCacheZoneMetric(namespace, "expired_responses", "Total number of cache hits with expired TTL", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"expired_bytes": newCacheZoneMetric(namespace, "expired_bytes", "Total number of bytes returned from cache hits with expired TTL", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"expired_responses_written": newCacheZoneMetric(namespace, "expired_responses_written", "Total number of cache hits with expired TTL written to cache", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"expired_bytes_written": newCacheZoneMetric(namespace, "expired_bytes_written", "Total number of bytes written to cache from cache hits with expired TTL", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"bypass_responses": newCacheZoneMetric(namespace, "bypass_responses", "Total number of cache bypasses", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"bypass_bytes": newCacheZoneMetric(namespace, "bypass_bytes", "Total number of bytes returned from cache bypasses", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"bypass_responses_written": newCacheZoneMetric(namespace, "bypass_responses_written", "Total number of cache bypasses written to cache", variableLabelNames.CacheZoneVariableLabelNames, constLabels),
"bypass_bytes_written": newCacheZoneMetric(namespace, "bypass_bytes_written", "Total number of bytes written to cache from cache bypasses", variableLabelNames.CacheZoneVariableLabelNames, constLabels),

Check warning on line 556 in collector/nginx_plus.go

View check run for this annotation

Codecov / codecov/patch

collector/nginx_plus.go#L536-L556 

Added lines #L536 - L556 were not covered by tests
},
workerMetrics: map[string]*prometheus.Desc{
"connection_accepted": newWorkerMetric(namespace, "connection_accepted", "The total number of accepted client connections", constLabels),
Expand Down Expand Up @@ -1193,34 +1200,39 @@
}

for name, zone := range stats.Caches {
var cold float64
if zone.Cold {
cold = 1.0
labelValues := []string{name}
varLabelValues := c.getCacheZoneLabelValues(name)

Check warning on line 1204 in collector/nginx_plus.go

View check run for this annotation

Codecov / codecov/patch

collector/nginx_plus.go#L1203-L1204 

Added lines #L1203 - L1204 were not covered by tests

if c.variableLabelNames.CacheZoneVariableLabelNames != nil && len(varLabelValues) != len(c.variableLabelNames.CacheZoneVariableLabelNames) {
level.Warn(c.logger).Log("msg", "wrong number of labels for cache zone, empty labels will be used instead", "zone", name, "labels", c.variableLabelNames.CacheZoneVariableLabelNames, "values", varLabelValues)
for range c.variableLabelNames.CacheZoneVariableLabelNames {
labelValues = append(labelValues, "")

Check warning on line 1209 in collector/nginx_plus.go

View check run for this annotation

Codecov / codecov/patch

collector/nginx_plus.go#L1206-L1209 

Added lines #L1206 - L1209 were not covered by tests
}
} else {
cold = 0.0
labelValues = append(labelValues, varLabelValues...)

Check warning on line 1212 in collector/nginx_plus.go

View check run for this annotation

Codecov / codecov/patch

collector/nginx_plus.go#L1212 

Added line #L1212 was not covered by tests
}

ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["size"], prometheus.GaugeValue, float64(zone.Size), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["max_size"], prometheus.GaugeValue, float64(zone.MaxSize), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["cold"], prometheus.GaugeValue, cold, name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["hit_responses"], prometheus.CounterValue, float64(zone.Hit.Responses), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["hit_bytes"], prometheus.CounterValue, float64(zone.Hit.Bytes), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["stale_responses"], prometheus.CounterValue, float64(zone.Stale.Responses), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["stale_bytes"], prometheus.CounterValue, float64(zone.Stale.Bytes), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["updating_responses"], prometheus.CounterValue, float64(zone.Updating.Responses), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["updating_bytes"], prometheus.CounterValue, float64(zone.Updating.Bytes), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["revalidated_responses"], prometheus.CounterValue, float64(zone.Revalidated.Responses), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["revalidated_bytes"], prometheus.CounterValue, float64(zone.Revalidated.Bytes), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["miss_responses"], prometheus.CounterValue, float64(zone.Miss.Responses), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["miss_bytes"], prometheus.CounterValue, float64(zone.Miss.Bytes), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["expired_responses"], prometheus.CounterValue, float64(zone.Expired.Responses), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["expired_bytes"], prometheus.CounterValue, float64(zone.Expired.Bytes), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["expired_responses_written"], prometheus.CounterValue, float64(zone.Expired.ResponsesWritten), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["expired_bytes_written"], prometheus.CounterValue, float64(zone.Expired.BytesWritten), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["bypass_responses"], prometheus.CounterValue, float64(zone.Bypass.Responses), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["bypass_bytes"], prometheus.CounterValue, float64(zone.Bypass.Bytes), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["bypass_responses_written"], prometheus.CounterValue, float64(zone.Bypass.ResponsesWritten), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["bypass_bytes_written"], prometheus.CounterValue, float64(zone.Bypass.BytesWritten), name)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["size"], prometheus.GaugeValue, float64(zone.Size), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["max_size"], prometheus.GaugeValue, float64(zone.MaxSize), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["cold"], prometheus.GaugeValue, booleanToFloat64[zone.Cold], labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["hit_responses"], prometheus.CounterValue, float64(zone.Hit.Responses), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["hit_bytes"], prometheus.CounterValue, float64(zone.Hit.Bytes), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["stale_responses"], prometheus.CounterValue, float64(zone.Stale.Responses), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["stale_bytes"], prometheus.CounterValue, float64(zone.Stale.Bytes), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["updating_responses"], prometheus.CounterValue, float64(zone.Updating.Responses), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["updating_bytes"], prometheus.CounterValue, float64(zone.Updating.Bytes), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["revalidated_responses"], prometheus.CounterValue, float64(zone.Revalidated.Responses), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["revalidated_bytes"], prometheus.CounterValue, float64(zone.Revalidated.Bytes), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["miss_responses"], prometheus.CounterValue, float64(zone.Miss.Responses), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["miss_bytes"], prometheus.CounterValue, float64(zone.Miss.Bytes), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["expired_responses"], prometheus.CounterValue, float64(zone.Expired.Responses), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["expired_bytes"], prometheus.CounterValue, float64(zone.Expired.Bytes), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["expired_responses_written"], prometheus.CounterValue, float64(zone.Expired.ResponsesWritten), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["expired_bytes_written"], prometheus.CounterValue, float64(zone.Expired.BytesWritten), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["bypass_responses"], prometheus.CounterValue, float64(zone.Bypass.Responses), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["bypass_bytes"], prometheus.CounterValue, float64(zone.Bypass.Bytes), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["bypass_responses_written"], prometheus.CounterValue, float64(zone.Bypass.ResponsesWritten), labelValues...)
ch <- prometheus.MustNewConstMetric(c.cacheZoneMetrics["bypass_bytes_written"], prometheus.CounterValue, float64(zone.Bypass.BytesWritten), labelValues...)

Check warning on line 1235 in collector/nginx_plus.go

View check run for this annotation

Codecov / codecov/patch

collector/nginx_plus.go#L1215-L1235 

Added lines #L1215 - L1235 were not covered by tests
}

for id, worker := range stats.Workers {
Expand All @@ -1244,6 +1256,11 @@
"unhealthy": 6.0,
}

var booleanToFloat64 = map[bool]float64{
true: 1.0,
false: 0.0,
}

func newServerZoneMetric(namespace string, metricName string, docString string, variableLabelNames []string, constLabels prometheus.Labels) *prometheus.Desc {
labels := []string{"server_zone"}
labels = append(labels, variableLabelNames...)
Expand Down Expand Up @@ -1304,8 +1321,10 @@
return prometheus.NewDesc(prometheus.BuildFQName(namespace, "stream_limit_connection", metricName), docString, []string{"zone"}, constLabels)
}

func newCacheZoneMetric(namespace string, metricName string, docString string, constLabels prometheus.Labels) *prometheus.Desc {
return prometheus.NewDesc(prometheus.BuildFQName(namespace, "cache", metricName), docString, []string{"zone"}, constLabels)
func newCacheZoneMetric(namespace string, metricName string, docString string, variableLabelNames []string, constLabels prometheus.Labels) *prometheus.Desc {
labels := []string{"zone"}
labels = append(labels, variableLabelNames...)
return prometheus.NewDesc(prometheus.BuildFQName(namespace, "cache", metricName), docString, labels, constLabels)

Check warning on line 1327 in collector/nginx_plus.go

View check run for this annotation

Codecov / codecov/patch

collector/nginx_plus.go#L1324-L1327 

Added lines #L1324 - L1327 were not covered by tests
}

func newWorkerMetric(namespace string, metricName string, docString string, constLabels prometheus.Labels) *prometheus.Desc {
Expand Down
Loading