From efb575284b41074ab41b5ecd1f12c2ac9231bac4 Mon Sep 17 00:00:00 2001
From: Matt Pryor <matt@stackhpc.com>
Date: Wed, 24 Jan 2024 15:20:13 +0000
Subject: [PATCH] Increase etcd timeouts (#227)
---
charts/openstack-cluster/values.yaml | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/charts/openstack-cluster/values.yaml b/charts/openstack-cluster/values.yaml
index c29e42e..1d84fa2 100644
--- a/charts/openstack-cluster/values.yaml
+++ b/charts/openstack-cluster/values.yaml
@@ -110,7 +110,14 @@ etcd:
dataDir: /var/lib/etcd
# Any extra command line arguments to pass to etcd
extraArgs:
- # Tell etcd to listen for metrics on 0.0.0.0 so Prometheus can collect them
+ # Set timeouts so that etcd tolerates 'slowness' (network + disks) better
+ # This is at the expense of taking longer to detect a leader failure
+ # https://etcd.io/docs/v3.5/tuning/#time-parameters
+ heartbeat-interval: "500" # defaults to 100ms in etcd 3.5
+ election-timeout: "5000" # defaults to 1000ms in etcd 3.5
+ # Set a slightly larger space quota than the default (default is 2GB)
+ quota-backend-bytes: "4294967296"
+ # Listen for metrics on 0.0.0.0 so Prometheus can collect them
listen-metrics-urls: http://0.0.0.0:2381
# The block device configuration for etcd
# If not specified, the root device is used