K3s SQLite Datastore

https://docs.k3s.io/datastore

Stores

SQLite

Sqlite Size (default)

du -sh /var/lib/rancher/k3s/server/db/*

Working Example:

8.0K    /var/lib/rancher/k3s/server/db/etcd
71M     /var/lib/rancher/k3s/server/db/state.db
32K     /var/lib/rancher/k3s/server/db/state.db-shm
4.3M    /var/lib/rancher/k3s/server/db/state.db-wal

As long as:

compaction is working,
and you restart k3s periodically the database should stay pretty small.

SQLITE vacuum

Recent releases of kine/k3s vacuum the sqlite database on startup. kine will also run a vacuum at startup,

If a manual vacuum reduces your size, k3s may be too old

SQLite Analyze with sqlite3_analyzer

apt-get update
apt-get install sqlite3
apt-get install sqlite3-tools
sqlite3_analyzer /var/lib/rancher/k3s/server/db/state.db

SQLite Report analytics on kine

Kine is an etcd shim that translates the etcd API to:

SQLite
Postgres
MySQL/MariaDB
NATS

sqlite3 /var/lib/rancher/k3s/server/db/state.db << EOF
.header on
.mode column
.dbinfo

SELECT
  COUNT(*) AS rows,
  (SELECT prev_revision FROM kine WHERE name = "compact_rev_key" LIMIT 1) AS compact_rev,
  MAX(id) AS current_rev
FROM kine;

SELECT COUNT(*), name FROM kine GROUP BY name ORDER BY COUNT(*) DESC LIMIT 50;

EOF

output

rows    compact_rev  current_rev
------  -----------  -----------
919771  44903333     45821272

COUNT(*)  name
--------  ------------------------------------------------------------
278873    /registry/leases/cert-manager/trust-manager-leader-election
77684     /registry/masterleases/188.245.43.250
71997     /registry/leases/kube-system/apiserver-fohhc347juam3arofccezbofla
71962     /registry/leases/kube-node-lease/kube-server-01.eraldy.com
49956     /registry/leases/kube-system/cert-manager-controller
49921     /registry/leases/kube-system/cert-manager-cainjector-leader-election
5229      /registry/cronjobs/mail/mail-checker-cron
4338      /registry/argoproj.io/applications/argocd/com-datacadamia
...

Etcd (not default)

Size

du -sh /var/lib/rancher/k3s/server/db/etcd/
# 8.0K    /var/lib/rancher/k3s/server/db/etcd/

Compaction

Compact is the process of deleting old rows from the database. The query that handles this is executed periodically by kine. See

Check your k3s journald logs for message that look like these, once every 5 minutes:

journalctl -e -u k3s --since "15 minutes ago" | grep -i compact

Oct 01 08:16:36 kube-server-01.eraldy.com k3s[2237448]: time="2025-10-01T08:16:36Z" level=info msg="COMPACT deleted 1000 rows from 1000 revisions in 906.760433ms - compacted to 43882333/45820240"
Oct 01 08:16:36 kube-server-01.eraldy.com k3s[2237448]: time="2025-10-01T08:16:36Z" level=info msg="COMPACT compactRev=43882333 targetCompactRev=43883333 currentRev=45820241"

Support

Manual Vacuum and Compaction

# check actual size
du -sh /var/lib/rancher/k3s/server/db/*
# ls -l /var/lib/rancher/k3s/server/db/
# drwx------ 2 root root        4096 Jun 24  2024 etcd
#-rw-r--r-- 1 root root 12666449920 Oct  1 08:38 state.db
#-rw-r--r-- 1 root root    17989632 Oct  1 08:42 state.db-shm
#-rw-r--r-- 1 root root  9257153872 Oct  1 08:41 state.db-wal

# Stop k3s
systemctl stop k3s

# Backup current database if enough disk space
cp /var/lib/rancher/k3s/server/db/state.db /root/state.db.backup

# Compact using sqlite
sqlite3 /var/lib/rancher/k3s/server/db/state.db 'VACUUM;'

# Check new size
du -sh /var/lib/rancher/k3s/server/db/*
# 8.0K    /var/lib/rancher/k3s/server/db/etcd
# 5.1G    /var/lib/rancher/k3s/server/db/state.db
# 64K     /var/lib/rancher/k3s/server/db/state.db-shm
# 32M     /var/lib/rancher/k3s/server/db/state.db-wal

Compaction (https://github.com/k3s-io/kine/issues/213#issuecomment-1717143914)

sqlite3 /var/lib/rancher/k3s/server/db/state.db << EOF
.header on
.mode column

delete from kine where id in (select id from (select id, name from kine where id not in (select max(id) as id from kine group by name)));

EOF

# Restart k3s
systemctl start k3s

Discussion

https://github.com/k3s-io/k3s/discussions/13011