feat: mixin, allow overriding of some labels by parameterizing mixin …

…recording/alert rules (#11495) Signed-off-by: Alexander Soelberg Heidarsson <[email protected]>
grafana · Oct 2, 2024 · f1425b6 · f1425b6
1 parent 8b4f095
commit f1425b6
Show file tree

Hide file tree

Showing 5 changed files with 29 additions and 14 deletions.
diff --git a/production/loki-mixin/alerts.libsonnet b/production/loki-mixin/alerts.libsonnet
@@ -6,12 +6,12 @@
         rules: [
           {
             alert: 'LokiRequestErrors',
-            expr: std.strReplace(|||
-              100 * sum(rate(loki_request_duration_seconds_count{status_code=~"5.."}[2m])) by (cluster, namespace, job, route)
+            expr: |||
+              100 * sum(rate(loki_request_duration_seconds_count{status_code=~"5.."}[2m])) by (%(group_by_cluster)s, job, route)
                 /
-              sum(rate(loki_request_duration_seconds_count[2m])) by (cluster, namespace, job, route)
+              sum(rate(loki_request_duration_seconds_count[2m])) by (%(group_by_cluster)s, job, route)
                 > 10
-            |||, 'cluster', $._config.per_cluster_label),
+            ||| % $._config,
             'for': '15m',
             labels: {
               severity: 'critical',
@@ -26,8 +26,8 @@
           {
             alert: 'LokiRequestPanics',
             expr: |||
-              sum(increase(loki_panic_total[10m])) by (%s, namespace, job) > 0
-            ||| % $._config.per_cluster_label,
+              sum(increase(loki_panic_total[10m])) by (%(group_by_cluster)s, job) > 0
+            ||| % $._config,
             labels: {
               severity: 'critical',
             },
@@ -41,8 +41,8 @@
           {
             alert: 'LokiRequestLatency',
             expr: |||
-              %s_namespace_job_route:loki_request_duration_seconds:99quantile{route!~"(?i).*tail.*|/schedulerpb.SchedulerForQuerier/QuerierLoop"} > 1
-            ||| % $._config.per_cluster_label,
+              %(group_prefix_jobs)s_route:loki_request_duration_seconds:99quantile{route!~"(?i).*tail.*|/schedulerpb.SchedulerForQuerier/QuerierLoop"} > 1
+            ||| % $._config,
             'for': '15m',
             labels: {
               severity: 'critical',
@@ -57,8 +57,8 @@
           {
             alert: 'LokiTooManyCompactorsRunning',
             expr: |||
-              sum(loki_boltdb_shipper_compactor_running) by (%s, namespace) > 1
-            ||| % $._config.per_cluster_label,
+              sum(loki_boltdb_shipper_compactor_running) by (%(group_by_cluster)s) > 1
+            ||| % $._config,
             'for': '5m',
             labels: {
               severity: 'warning',

diff --git a/production/loki-mixin/config.libsonnet b/production/loki-mixin/config.libsonnet
@@ -1,4 +1,7 @@
 {
+  local makePrefix(groups) = std.join('_', groups),
+  local makeGroupBy(groups) = std.join(', ', groups),
+
   _config+:: {
     // Tags for dashboards.
     tags: ['loki'],
@@ -11,6 +14,20 @@
 
     // The label used to differentiate between different clusters.
     per_cluster_label: 'cluster',
+    per_namespace_label: 'namespace',
+    per_job_label: 'job',
+
+    // Grouping labels, to uniquely identify and group by {jobs, clusters}
+    job_labels: [$._config.per_cluster_label, $._config.per_namespace_label, $._config.per_job_label],
+    cluster_labels: [$._config.per_cluster_label, $._config.per_namespace_label],
+
+    // Each group prefix is composed of `_`-separated labels
+    group_prefix_jobs: makePrefix($._config.job_labels),
+    group_prefix_clusters: makePrefix($._config.cluster_labels),
+
+    // Each group-by label list is `, `-separated and unique identifies
+    group_by_job: makeGroupBy($._config.job_labels),
+    group_by_cluster: makeGroupBy($._config.cluster_labels),
 
     // Enable dashboard and panels for Grafana Labs internal components.
     internal_components: false,

diff --git a/production/loki-mixin/dashboards.libsonnet b/production/loki-mixin/dashboards.libsonnet
@@ -1,4 +1,3 @@
-(import 'config.libsonnet') +
 (import 'dashboards/loki-retention.libsonnet') +
 (import 'dashboards/loki-chunks.libsonnet') +
 (import 'dashboards/loki-logs.libsonnet') +

diff --git a/production/loki-mixin/mixin-ssd.libsonnet b/production/loki-mixin/mixin-ssd.libsonnet
@@ -1,6 +1,4 @@
-(import 'dashboards.libsonnet') +
-(import 'alerts.libsonnet') +
-(import 'recording_rules.libsonnet') + {
+(import 'mixin.libsonnet') + {
   grafanaDashboardFolder: 'Loki SSD',
 
   _config+:: {

diff --git a/production/loki-mixin/mixin.libsonnet b/production/loki-mixin/mixin.libsonnet
@@ -1,5 +1,6 @@
 (import 'dashboards.libsonnet') +
 (import 'alerts.libsonnet') +
+(import 'config.libsonnet') +
 (import 'recording_rules.libsonnet') + {
   grafanaDashboardFolder: 'Loki',
   // Without this, configs is not taken into account