You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I setup a new etcd cluster and for some reason each node is reporting 2 distinctetcd_cluster_version (v3.0 and v3.4) at the same time even though each node is reporting 1 etcd_server_version.
etcd_cluster_version:
3.4.0 | 10.10.10.33:2479 | Up
3.0.0 | 10.10.10.33:2479 | Up
3.4.0 | 10.10.10.92:2479 | Up
3.0.0 | 10.10.10.92:2479 | Up
3.4.0 | 10.10.10.15:2479 | Up
3.0.0 | 10.10.10.15:2479 | Up
3.4.0 | 10.10.10.241:2479 | Up
3.0.0 | 10.10.10.241:2479 | Up
3.4.0 | 10.10.10.190:2479 | Up
3.0.0 | 10.10.10.190:2479 | Up
[...]
etcd_server_version:
10.10.10.241:2479 | 3.4.0 | Up
10.10.10.190:2479 | 3.4.0 | Up
10.10.10.15:2479 | 3.4.0 | Up
10.10.10.247:2479 | 3.4.0 | Up
10.10.10.33:2479 | 3.4.0 | Up
10.10.10.92:2479 | 3.4.0 | Up
10.10.10.167:2479 | 3.4.0 | Up
I double checked on some hosts: it's not an issue with the metrics acquisition:
root@etcd-dasiuod:~# curl http://127.0.0.1:2479/metrics
# HELP etcd_cluster_version Which version is running. 1 for 'cluster_version' label with current cluster version
# TYPE etcd_cluster_version gauge
etcd_cluster_version{cluster_version="3.0.0"} 1
etcd_cluster_version{cluster_version="3.4.0"} 1
[...]
Note that this is a brand new cluster, not an upgrade. We have other 3.4 etcd clusters that don't behave like that. Any idea? Thanks!
The text was updated successfully, but these errors were encountered:
After investigation it seems that this is being caused by this function: https://github.com/etcd-io/etcd/blob/release-3.4/etcdserver/api/membership/cluster.go#L541 .
When the cluster is initially started, it starts off at 3.0: etcdserver/membership: set the initial cluster version to 3.0
And as a result the metric etcd_cluster_version{cluster_version="3.0.0"} is set to 1. But then, the cluster version is updated to 3.4: etcdserver/membership: updated the cluster version from 3.0 to 3.4
So as a result the metric etcd_cluster_version{cluster_version="3.4.0"} is also set to 1. But in this logic, etcd_cluster_version{cluster_version="3.0.0"} is never set back to 0, so the metrics (etcd_cluster_version} show it as running 2 versions.
I setup a new etcd cluster and for some reason each node is reporting 2 distinct
etcd_cluster_version
(v3.0
andv3.4
) at the same time even though each node is reporting 1etcd_server_version
.etcd_cluster_version:
etcd_server_version:
I double checked on some hosts: it's not an issue with the metrics acquisition:
Note that this is a brand new cluster, not an upgrade. We have other
3.4
etcd clusters that don't behave like that. Any idea? Thanks!The text was updated successfully, but these errors were encountered: