Backup failed due to TooManySnapshots error

Description

During backup, Longhorn volumes are backed up by taking the volume snapshot and sending it to a remote location. If snapshot creation issues affect the Longhorn volume, with the snapshot count for the volume being higher than 248, the backup will not succeed.

You can verify if the backup failed due to the TooManySnapshots error in the following ways:

• By checking the Velero logs:

  kubectl logs  -n velero -l app.kubernetes.io/name=velero -c velero |grep "Waiting for volumesnapshotcontents"

  Sample output:

  time="2023-12-15T08:15:59Z" level=info msg="Waiting for volumesnapshotcontents snapcontent-f073b88e-bd01-40a8-a645-ec929e276cef to have snapshot handle. Retrying in 5s" backup=velero/daily-2 cmd=/plugins/velero-plugin-for-csi logSource="/go/src/velero-plugin-for-csi/internal/util/util.go:182" pluginName=velero-plugin-for-csi
  time="2023-12-15T08:15:59Z" level=info msg="Waiting for volumesnapshotcontents snapcontent-f073b88e-bd01-40a8-a645-ec929e276cef to have snapshot handle. Retrying in 5s" backup=velero/daily-2 cmd=/plugins/velero-plugin-for-csi logSource="/go/src/velero-plugin-for-csi/internal/util/util.go:182" pluginName=velero-plugin-for-csi

• By checking the Longhorn pod logs:

  kubectl logs  -n longhorn-system -l app=csi-snapshotter --tail=-1 |grep "too many snapshots created"

  Sample output:

  I1215 08:39:41.707351       1 snapshot_controller.go:291] createSnapshotWrapper: CreateSnapshot for content snapcontent-f073b88e-bd01-40a8-a645-ec929e276cef returned error: rpc error: code = Internal desc = Bad response statusCode [500]. Status [500 Internal Server Error]. Body: [code=Server Error, detail=, message=failed to create snapshot: proxyServer=10.42.7.56:8501 destination=10.42.7.56:10004: failed to snapshot volume: rpc error: code = Unknown desc = failed to create snapshot snapshot-f073b88e-bd01-40a8-a645-ec929e276cef for volume 10.42.7.56:10004: rpc error: code = Unknown desc = too many snapshots created] from [http://longhorn-backend:9500/v1/volumes/pvc-7d89efa4-3d60-4837-a632-f190cd3cd9ed?action=snapshotCreate]

Solution

If you encounter the TooManySnapshots error, you must clean up the snapshots for all the volumes by running the following script. For details on how to automate this operation, see How to automatically clean up Longhorn snapshots.

#!/bin/bash
set -e

# longhorn backend URL
url=
# By default, snapshot older than 10 days will be deleted
days=10

function display_usage() {
	echo "usage: $(basename "$0") [-h] -u longhorn-url -d days"
	echo "  -u	Longhorn URL"
	echo "  -d 	Number of days(should be >0). By default, script will delete snapshot older than 10 days."
	echo "  -h	Print help"
}

while getopts 'hd:u:' flag "$@"; do
	case "${flag}" in
		u)
			url=${OPTARG}
			;;
		d)
			days=${OPTARG}
			[ "$days" ] && [ -z "${days//[0-9]}" ] || { echo "Invalid number of days=$days"; exit 1; }
			;;
		h)
			display_usage
			exit 0
			;;
		:)
			echo "Invalid option: ${OPTARG} requires an argument."
			exit 1
			;;
		*)
			echo "Unexpected option ${flag}"
			exit 1
			;;
	esac
done

[[ -z "$url" ]] && echo "Missing longhorn URL" && exit 1

# check if URL is valid
curl -s --connect-timeout 30 ${url}/v1 >> /dev/null || { echo "Unable to connect to longhorn backend"; exit 1; }

echo "Deleting snapshots older than $days days"

# Fetch list of longhorn volumes
vols=$( (curl -s -X GET ${url}/v1/volumes |jq -r '.data[].name') )

#delete given snapshot for given volume
function delete_snapshot() {
	local vol=$1
	local snap=$2

	[[ -z "$vol" || -z "$snap" ]] && echo "Error: delete_snapshot: Empty argument" && return 1
	curl -s -X POST ${url}/v1/volumes/${vol}?action=snapshotDelete -d '{"name": "'$snap'"}'
	echo "Snapshot=$snap deleted for volume=$vol"
}

#perform cleanup for given volume
function cleanup_volume() {
	local vol=$1
	local deleted_snap=0

	[[ -z "$vol" ]] && echo "Error: cleanup_volume: Empty argument" && return 1

	# fetch list of snapshot
	snaps=$( (curl -s -X POST ${url}/v1/volumes/${vol}?action=snapshotList | jq  -r '.data[] | select(.usercreated==true) | .name' ) )
	for i in ${snaps[@]}; do
		if [[ $i == "volume-head" ]]; then
			continue
		fi

		# calculate date difference for snapshot
		snapTime=$(curl -s -X POST ${url}/v1/volumes/${vol}?action=snapshotGet -d '{"name":"'$i'"}' |jq -r '.created')
		currentTime=$(date "+%s")
		timeDiff=$(($currentTime - ($(date -d $snapTime "+%s")) / 86400))
		if [[ $timeDiff -lt $days ]]; then
			echo "Ignoring snapshot $i, since it is older than $timeDiff days"
			continue
		fi

		#trigger deletion for snapshot
		delete_snapshot $vol $i
		deleted_snap=$((deleted_snap+1))
	done

	if [[ "$deleted_snap" -gt 0 ]]; then
		#trigger purge for volume
		curl -s -X POST ${url}/v1/volumes/${vol}?action=snapshotPurge >> /dev/null
	fi

}

for i in ${vols[@]}; do
	cleanup_volume $i
done

When executing the aforementioned script, you must pass the following arguments:

• -u - The Longhorn backend URL. To get the Longhorn backend URL, run the following command:

  kubectl get svc -n longhorn-system longhorn-backend -o json | jq -r '.spec | (.clusterIP|tostring) + ":" + (.ports[0].port|tostring)' to fetch URL

• -d - The number of days before the snapshots are deleted.

Note:

To get a list of volumes affected by the TooManySnapshots error, run the following command:

kubectl get volumes -n longhorn-system -o json | jq -r '.items[] | select(([ .status.conditions[] | select(.type == "toomanysnapshots" and .status == "True") ] | length ) == 1 ) | .metadata.name'