bstein/titan-iac
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

longhorn: ensure engine image on oceanus

Browse Source
This commit is contained in:
jenkins 2026-06-09 02:06:34 -03:00
parent 8fcc61cae9
commit 63c869bf42
1 changed files with 44 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

58
infrastructure/longhorn/core/longhorn-csi-toleration-ensure-job.yaml
View File

@ -2,7 +2,7 @@
apiVersion: batch/v1
kind: Job
metadata:
name: longhorn-csi-toleration-ensure-3
name: longhorn-csi-toleration-ensure-4
namespace: longhorn-system
spec:
backoffLimit: 0
@ -38,14 +38,18 @@ spec:
value="true"
effect="NoSchedule"
current="$(kubectl -n "${ns}" get daemonset "${ds}" -o json)"
if echo "${current}" | jq -e \
--arg key "${key}" \
--arg value "${value}" \
--arg effect "${effect}" \
'.spec.template.spec.tolerations[]? | select(.key == $key and .value == $value and .effect == $effect)' >/dev/null; then
echo "${ds} already tolerates ${key}=${value}:${effect}"
else
patch_daemonset() {
target="$1"
current="$(kubectl -n "${ns}" get daemonset "${target}" -o json)"
if echo "${current}" | jq -e \
--arg key "${key}" \
--arg value "${value}" \
--arg effect "${effect}" \
'.spec.template.spec.tolerations[]? | select(.key == $key and .value == $value and .effect == $effect)' >/dev/null; then
echo "${target} already tolerates ${key}=${value}:${effect}"
return 0
fi
patch="$(echo "${current}" | jq -c \
--arg key "${key}" \
--arg value "${value}" \
@ -61,16 +65,42 @@ spec:
}
}
}')"
kubectl -n "${ns}" patch daemonset "${ds}" --type=merge -p "${patch}"
fi
kubectl -n "${ns}" patch daemonset "${target}" --type=merge -p "${patch}"
}
patch_daemonset "${ds}"
engine_daemonsets="$(kubectl -n "${ns}" get daemonset -l longhorn.io/component=engine-image -o jsonpath='{range .items[*]}{.metadata.name}{"\n"}{end}')"
for engine_ds in ${engine_daemonsets}; do
patch_daemonset "${engine_ds}"
done
csi_ready="false"
for attempt in $(seq 1 90); do
if kubectl get csinode titan-23 -o json | jq -e '.spec.drivers[]? | select(.name == "driver.longhorn.io")' >/dev/null; then
echo "driver.longhorn.io registered on titan-23"
exit 0
csi_ready="true"
break
fi
sleep 2
done
echo "driver.longhorn.io did not register on titan-23 before timeout" >&2
exit 1
if [ "${csi_ready}" != "true" ]; then
echo "driver.longhorn.io did not register on titan-23 before timeout" >&2
exit 1
fi
for engine_ds in ${engine_daemonsets}; do
for attempt in $(seq 1 90); do
if kubectl -n "${ns}" get pods -o json | jq -e \
--arg engine_ds "${engine_ds}" \
'.items[] | select(.spec.nodeName == "titan-23") | select(.metadata.ownerReferences[]?.name == $engine_ds) | select([.status.containerStatuses[]?.ready] | all)' >/dev/null; then
echo "${engine_ds} ready on titan-23"
break
fi
if [ "${attempt}" = "90" ]; then
echo "${engine_ds} did not become ready on titan-23 before timeout" >&2
exit 1
fi
sleep 2
done
done