ppetru/alo-cluster
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Only keep 10 snapshots, and push metrics.

Browse Source
This commit is contained in:
Petru
2025-11-04 10:22:11 +00:00
parent 49afc0c084
commit 5ef4d832fb
2 changed files with 40 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
common/nfs-services-standby.nix
View File

@@ -39,26 +39,27 @@ in
noCheck = true;
};
# Cleanup old snapshots on standby (keep last 4 hours for HA failover)
# Cleanup old snapshots on standby (keep last 10 snapshots)
systemd.services.cleanup-services-standby-snapshots = {
description = "Cleanup old btrfs snapshots in services-standby";
path = [ pkgs.btrfs-progs pkgs.findutils pkgs.coreutils ];
path = [ pkgs.btrfs-progs pkgs.findutils pkgs.coreutils pkgs.curl ];
script = ''
set -euo pipefail
# Keep at least 2 hours of snapshots (24 snapshots at 5min intervals)
MIN_KEEP=24
# Cleanup old snapshots on standby (keep last 10 snapshots, sorted by name/timestamp)
ls -1d /persist/services-standby/services@* 2>/dev/null | sort | head -n -10 | xargs -r btrfs subvolume delete || true
# Count existing snapshots
count=$(find /persist/services-standby -maxdepth 1 -name 'services@*' -type d | wc -l)
# Calculate metrics
CLEANUP_TIME=$(date +%s)
SNAPSHOT_COUNT=$(ls -1d /persist/services-standby/services@* 2>/dev/null | wc -l)
# Only delete old snapshots if we have more than the minimum
if [ $count -gt $MIN_KEEP ]; then
# Delete snapshots older than 4 hours
find /persist/services-standby -maxdepth 1 -name 'services@*' -mmin +240 -exec btrfs subvolume delete {} \; || true
else
echo "Only $count snapshots found, keeping all (minimum: $MIN_KEEP)"
fi
# Push metrics to Prometheus pushgateway
cat <<METRICS | curl --data-binary @- http://pushgateway.service.consul:9091/metrics/job/nfs_standby_cleanup/instance/$(hostname)
# TYPE nfs_standby_snapshot_count gauge
nfs_standby_snapshot_count $SNAPSHOT_COUNT
# TYPE nfs_standby_cleanup_last_run_timestamp gauge
nfs_standby_cleanup_last_run_timestamp $CLEANUP_TIME
METRICS
'';
serviceConfig = {
Type = "oneshot";