(couchdb) branch main updated: Add detailed node membership info to prometheus

[vatamane]

This is an automated email from the ASF dual-hosted git repository.

vatamane pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/couchdb.git


The following commit(s) were added to refs/heads/main by this push:
     new 84ed47f91 Add detailed node membership info to prometheus
84ed47f91 is described below

commit 84ed47f916cbeb652f3bccf7a46db6b73b00e7cf
Author: Nick Vatamaniuc <vatam...@gmail.com>
AuthorDate: Fri Apr 25 19:16:27 2025 -0400

    Add detailed node membership info to prometheus
    
    One of our users in #5485 reported they would like to be able have a more
    detailed picture of node connections and disconnections (are expected nodes
    connected or disconnected) in prometheus, so that's what this PR does.
    
    We already used per-node metrics for dist channel stats so went with the 
same
    format: `{node="..."} 0/1`
    
    ```
    couchdb_membership{nodes="cluster_nodes"} 3
    couchdb_membership{nodes="all_nodes"} 3
    couchdb_membership{node="node2@127.0.0.1"} 1
    couchdb_membership{node="node3@127.0.0.1"} 1
    
    couchdb_erlang_distribution_recv_oct_bytes_total{node="node2@127.0.0.1"} 
40962
    couchdb_erlang_distribution_recv_oct_bytes_total{node="node3@127.0.0.1"} 
25681
    ```
    
    Fix: #5485
---
 src/couch_prometheus/src/couch_prometheus.erl | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/src/couch_prometheus/src/couch_prometheus.erl 
b/src/couch_prometheus/src/couch_prometheus.erl
index ee9a2b6ce..4d053d1af 100644
--- a/src/couch_prometheus/src/couch_prometheus.erl
+++ b/src/couch_prometheus/src/couch_prometheus.erl
@@ -67,6 +67,7 @@ get_system_stats() ->
         get_ets_stats(),
         get_internal_replication_jobs_stat(),
         get_membership_stat(),
+        get_membership_nodes(),
         get_distribution_stats()
     ]).
 
@@ -98,6 +99,19 @@ get_membership_stat() ->
     ],
     to_prom(membership, gauge, "count of nodes in the cluster", Labels).
 
+get_membership_nodes() ->
+    Self = config:node_name(),
+    Expected = mem3:nodes() -- [Self],
+    Nodes = nodes(),
+    Disconnected = Expected -- Nodes,
+    % Extra nodes, connected but not part of mem3
+    Extra = Nodes -- Expected,
+    Connected = Nodes -- Extra,
+    ConnectedLabels = [{[{node, N}], 1} || N <- lists:sort(Connected)],
+    DisconnectedLables = [{[{node, N}], 0} || N <- lists:sort(Disconnected)],
+    Labels = ConnectedLabels ++ DisconnectedLables,
+    to_prom(membership_nodes, gauge, "cluster node connectivity", Labels).
+
 get_vm_stats() ->
     MemLabels = lists:map(
         fun({Type, Value}) ->