gsadmin/rustfs
1
0
Fork 0
mirror of https://github.com/rustfs/rustfs.git synced 2026-08-11 23:56:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(helm): coordinate distributed pod startup (#5258)

Browse Source
This commit is contained in:
cxymds
2026-07-26 11:37:01 +08:00
committed by GitHub
parent bdf3f0484d
commit 50c4dcca4f
3 changed files with 103 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
helm/rustfs/templates/statefulset.yaml
+47 -1
View File
@@ -4,9 +4,13 @@
{{- if and .Values.mode.distributed.enabled (le (int .Values.replicaCount) 1) -}}
{{- fail "Distributed mode requires replicaCount >= 2" -}}
{{- end -}}
{{- if and .Values.mode.distributed.enabled (le (int .Values.startupWaitTimeoutSeconds) 0) -}}
{{- fail "startupWaitTimeoutSeconds must be greater than 0 in distributed mode" -}}
{{- end -}}
{{- if .Values.mode.distributed.enabled }}
{{- range $pool := include "rustfs.pools" . | fromJsonArray }}
{{- $pools := include "rustfs.pools" . | fromJsonArray }}
{{- range $pool := $pools }}
{{- $drivesPerNode := int $pool.drives }}
---
apiVersion: apps/v1
@@ -123,6 +127,10 @@ spec:
env:
- name: DRIVES_PER_NODE
value: {{ $drivesPerNode | quote }}
- name: ENDPOINT_PORT
value: {{ $.Values.service.endpoint.port | quote }}
- name: STARTUP_WAIT_TIMEOUT_SECONDS
value: {{ $.Values.startupWaitTimeoutSeconds | quote }}
command:
- sh
- -c
@@ -138,6 +146,44 @@ spec:
mkdir -p /mnt/rustfs/logs
chmod 755 /mnt/rustfs/logs
{{- end }}
deadline=$(($(date +%s) + STARTUP_WAIT_TIMEOUT_SECONDS))
wait_for_peer() {
description=$1
shift
until "$@" >/dev/null 2>&1; do
if [ "$(date +%s)" -ge "$deadline" ]; then
echo "Timed out after ${STARTUP_WAIT_TIMEOUT_SECONDS}s waiting for ${description}" >&2
exit 1
fi
sleep 2
done
}
echo "Waiting for distributed peer DNS records"
{{- range $peerPool := $pools }}
{{- range $i := until (int $peerPool.replicaCount) }}
wait_for_peer "DNS for {{ $peerPool.fullname }}-{{ $i }}" \
nslookup "{{ $peerPool.fullname }}-{{ $i }}.{{ include "rustfs.fullname" $ }}-headless.{{ $.Release.Namespace }}.svc.{{ include "rustfs.clusterDomain" $ }}"
{{- end }}
{{- end }}
ordinal=${HOSTNAME##*-}
echo "Waiting for earlier distributed peers to listen on port ${ENDPOINT_PORT}"
{{- range $peerPool := $pools }}
{{- if lt (int $peerPool.index) (int $pool.index) }}
{{- range $i := until (int $peerPool.replicaCount) }}
wait_for_peer "{{ $peerPool.fullname }}-{{ $i }}:${ENDPOINT_PORT}" \
nc -z -w 2 "{{ $peerPool.fullname }}-{{ $i }}.{{ include "rustfs.fullname" $ }}-headless.{{ $.Release.Namespace }}.svc.{{ include "rustfs.clusterDomain" $ }}" "$ENDPOINT_PORT"
{{- end }}
{{- else if eq (int $peerPool.index) (int $pool.index) }}
i=0
while [ "$i" -lt "$ordinal" ]; do
wait_for_peer "{{ $peerPool.fullname }}-${i}:${ENDPOINT_PORT}" \
nc -z -w 2 "{{ $peerPool.fullname }}-${i}.{{ include "rustfs.fullname" $ }}-headless.{{ $.Release.Namespace }}.svc.{{ include "rustfs.clusterDomain" $ }}" "$ENDPOINT_PORT"
i=$((i + 1))
done
{{- end }}
{{- end }}
volumeMounts:
{{- if gt $drivesPerNode 1 }}
{{- range $i := until $drivesPerNode }}
helm/rustfs/values.yaml
+5
View File
@@ -30,6 +30,11 @@ drivesPerNode: null
# Provide the DNS root only, without a `svc.` prefix or leading/trailing dots.
clusterDomain: cluster.local
# Maximum time an init container waits for distributed peers to publish DNS
# and for earlier peers to start listening. The wait is shared across all
# peers, not applied once per peer.
startupWaitTimeoutSeconds: 300
# This sets the container image more information can be found here: https://kubernetes.io/docs/concepts/containers/images/
image:
rustfs: # This sets the rustfs image repository and tag.