cgroup-rstat-flushing: Default disable per cgroup recording

Recording two counters per cgroup generates too much data for prometheus. For troubleshooting this will be a practical feature, but don't enable this on all servers per default. Signed-off-by: Jesper Dangaard Brouer <[email protected]>
cloudflare · Aug 19, 2024 · 3397837 · 3397837
1 parent ebe07e7
commit 3397837
Show file tree

Hide file tree

Showing 2 changed files with 35 additions and 24 deletions.
diff --git a/examples/cgroup-rstat-flushing.bpf.c b/examples/cgroup-rstat-flushing.bpf.c
@@ -86,6 +86,12 @@ struct flush_key_t {
     u32 level;
 } __attribute__((packed));
 
+/* This provide easy way to enable per cgroup flush recording
+ *  Default this is disabled, as it generates too much data.
+ *  For troubleshooting purposes this can be enabled on some servers.
+ */
+#undef CONFIG_TRACK_PER_CGROUP_FLUSH
+#ifdef CONFIG_TRACK_PER_CGROUP_FLUSH
 struct {
     __uint(type, BPF_MAP_TYPE_PERCPU_HASH);
     __uint(max_entries, 10000);
@@ -99,6 +105,7 @@ struct {
     __type(key, struct flush_key_t);
     __type(value, u64);
 } cgroup_rstat_flush_nanoseconds_count SEC(".maps");
+#endif /* CONFIG_TRACK_PER_CGROUP_FLUSH */
 
 /* Complex key for encoding lock properties */
 struct lock_key_t {
@@ -386,12 +393,14 @@ int BPF_PROG(cgroup_rstat_flush_locked_exit, struct cgroup *cgrp)
      *   rate(_seconds_sum[1m]) / rate(_seconds_count[1m])
      */
 
+#ifdef CONFIG_TRACK_PER_CGROUP_FLUSH
     /* Per cgroup record average latency via nanoseconds_count and nanoseconds_sum */
     struct flush_key_t flush_key;
     flush_key.cgrp_id = cgroup_id(cgrp);
     flush_key.level = cgrp->level;
     increment_map_nosync(&cgroup_rstat_flush_nanoseconds_sum, &flush_key, delta_ns);
     increment_map_nosync(&cgroup_rstat_flush_nanoseconds_count, &flush_key, 1);
+#endif
 
     return 0;
 }

diff --git a/examples/cgroup-rstat-flushing.yaml b/examples/cgroup-rstat-flushing.yaml
@@ -22,30 +22,32 @@ metrics:
           size: 2
           decoders:
             - name: uint
-    - name: cgroup_rstat_flush_nanoseconds_sum
-      help: Sum amount of time (in nanoseconds) spent for flushes per cgroup
-      labels:
-        - name: cgroup
-          size: 8
-          decoders:
-            - name: uint
-            - name: cgroup
-        - name: level
-          size: 4
-          decoders:
-            - name: uint
-    - name: cgroup_rstat_flush_nanoseconds_count
-      help: Number of flush calls related to cgroup_rstat_flush_nanoseconds_sum
-      labels:
-        - name: cgroup
-          size: 8
-          decoders:
-            - name: uint
-            - name: cgroup
-        - name: level
-          size: 4
-          decoders:
-            - name: uint
+# --- See: CONFIG_TRACK_PER_CGROUP_FLUSH
+#    - name: cgroup_rstat_flush_nanoseconds_sum
+#      help: Sum amount of time (in nanoseconds) spent for flushes per cgroup
+#      labels:
+#        - name: cgroup
+#          size: 8
+#          decoders:
+#            - name: uint
+#            - name: cgroup
+#        - name: level
+#          size: 4
+#          decoders:
+#            - name: uint
+#    - name: cgroup_rstat_flush_nanoseconds_count
+#      help: Number of flush calls related to cgroup_rstat_flush_nanoseconds_sum
+#      labels:
+#        - name: cgroup
+#          size: 8
+#          decoders:
+#            - name: uint
+#            - name: cgroup
+#        - name: level
+#          size: 4
+#          decoders:
+#            - name: uint
+# ---
     - name: cgroup_rstat_map_errors_total
       help: Map related errors
       labels: