Move the database to Postgres, before teams need the schema
CI / chart (pull_request) Successful in 1s
CI / security (pull_request) Successful in 17s
CI / test (pull_request) Successful in 2m5s

First step of #1, and it goes first for one reason: #4 adds a team_id to
nearly every table, and doing that twice -- once for SQLite, once for
Postgres -- is work nobody gets paid for. The teams migrations now only
have to be written against one database.

The ten SQLite migrations are replaced by a single Postgres baseline
rather than ported one by one. They were incremental in a way that has
no value on a fresh install: 004 adds columns 008 drops again, and 008's
backfill rewrites data a Postgres database never had. The history stays
in git; the schema they add up to is now 001_baseline.sql.

Timestamps stay BIGINT unix seconds and are NOT converted to timestamptz.
Everything in Go already speaks epochs, so converting would have been a
second, larger change riding along inside this one. It is worth doing on
its own. The JSON columns did move to jsonb, because #4 will want to
filter and index on labels.

Most of the port is mechanical -- 170 placeholders from ? to $1 -- but
four things needed more than a search and replace:

  * Dynamically built WHERE clauses cannot keep their numbering straight
    by hand, so they hand out placeholders through sqlArgs instead. A
    filter can now be added or reordered without renumbering anything.

  * SUM(resolved_at IS NULL) was SQLite counting a boolean as 0 or 1.
    Postgres has no sum(boolean), and this was breaking every dead man's
    switch -- silently, since the sweeper only logs. Now COUNT(*) FILTER.

  * unixepoch() became FLOOR(EXTRACT(EPOCH FROM now()))::bigint. The
    FLOOR is load-bearing: a bare cast rounds half up, so a row written
    at .6 of a second claimed a timestamp a second in the future and
    disagreed with the time.Now().Unix() the Go side stamps.

  * The unique-violation check matched SQLite's error text. It matches
    SQLSTATE 23505 now, so a renamed constraint cannot turn a 409 back
    into a 500.

Tests need a real Postgres, because there is no in-memory Postgres the
way there was an in-memory SQLite. Each test gets its own schema on a
shared server -- cheaper than a database each, and still isolated.
TERDUT_TEST_DSN says where it is; `make test-db` starts one locally and
ci.yaml runs one as a service container. An unset DSN fails the suite
rather than skipping it: a run that quietly tests nothing is worse than
one that does not run.

TestMigration_BackfillCarriesAckAndComments is deleted along with the
migrations it replayed. What it protected -- an upgrade not losing
acknowledgements and comments -- now belongs to scripts/sqlite-to-postgres.go,
which is build-tagged so the SQLite driver stays out of the server
binary. Both are meant to be deleted once this install has migrated.

The chart loses the PVC, the data volume and the python backup sidecar,
and requires database.dsnSecret.name: it provisions no database and
cannot guess where the credentials live, so a render without it is meant
to fail. Backups move to where Postgres actually runs. The other half of
that -- the postgresql CR, the k8up pg_dump annotation and the network
policy -- is a change to the wrapper chart in Ryuvia/charts and is not in
here.

Verified rather than assumed: the gate is green with -race against
Postgres 17, govulncheck and gitleaks are clean, and the migration script
was run end to end against a SQLite database built at the old schema and
seeded in every table. Ids survive, so incidents keep their numbers and
every foreign key still points where it did; the identity sequences are
moved past the copied ids, and a webhook after the migration opened
incident 12 rather than colliding at 1.
This commit is contained in:
Niklas Ye
2026-09-20 10:44:12 +02:00
parent 989425e550
commit dc39e3a5d3
44 changed files with 1004 additions and 725 deletions
+12 -65
View File
@@ -10,50 +10,15 @@ spec:
selector:
matchLabels:
{{- include "terdut-server.selectorLabels" . | nindent 6 }}
# The data PVC is ReadWriteOnce, so a RollingUpdate deadlocks: the new pod
# cannot attach the volume until the old one releases it, and the old one is
# not torn down until the new one is ready.
# Recreate, not RollingUpdate, even though the PVC that forced it is gone: the
# sweeper and the notifier are unsynchronised singletons, and two replicas
# overlapping during a rollout would both page for the same incident.
strategy:
type: Recreate
template:
metadata:
labels:
{{- include "terdut-server.selectorLabels" . | nindent 8 }}
{{- if .Values.backupSidecar.enabled }}
annotations:
# Dumps the whole database: incidents, alerts, users, API key hashes,
# the schedule and the notification outbox.
#
# Runs in the `backup` sidecar, NOT in the app container: the server
# image is FROM scratch and has no interpreter at all. k8up execs into
# .spec.containers[0] unless told otherwise, hence the explicit
# k8up.io/backupcommand-container.
#
# Buffered and sanity-checked before the first byte reaches stdout: k8up
# streams stdout straight into restic, so a dump that dies partway is
# stored as a silently-truncated snapshot that k8up still reports as
# Succeeded. The check counts users rather than incidents -- incidents
# are swept and archived, so an empty incidents table is a legitimate
# state, whereas a database with no users never is.
#
# The connection is read-only but the mount is not: the database runs in
# WAL mode, and opening it mode=ro still needs write access to the -shm
# wal-index.
#
# chr(10), not '\n': k8up parses this annotation with go-shellquote.
k8up.io/backupcommand-container: backup
k8up.io/backupcommand: >-
python3 -c "import sqlite3, sys;
con = sqlite3.connect('file:/data/terdut.db?mode=ro', uri=True);
con.execute('BEGIN');
users = con.execute('SELECT count(*) FROM users').fetchone()[0];
out = chr(10).join(con.iterdump()) + chr(10);
(users > 0 and out.rstrip().endswith('COMMIT;'))
or sys.exit('terdut: db dump failed sanity checks');
sys.stdout.write(out)"
k8up.io/file-extension: ".sql"
k8up.io/backup: "true"
{{- end }}
spec:
enableServiceLinks: false
containers:
@@ -67,8 +32,14 @@ spec:
env:
- name: TERDUT_ADDR
value: ":{{ .Values.service.port }}"
- name: TERDUT_DB_PATH
value: "/data/terdut.db"
# The connection string, from a Secret: it carries the password.
# The wrapper chart points this at the Secret the Postgres operator
# writes for this database's role.
- name: TERDUT_DB_DSN
valueFrom:
secretKeyRef:
name: {{ required "database.dsnSecret.name is required" .Values.database.dsnSecret.name }}
key: {{ .Values.database.dsnSecret.key }}
- name: TERDUT_STALE_AFTER
value: "{{ .Values.sweeper.staleAfter }}"
- name: TERDUT_ARCHIVE_AFTER
@@ -96,9 +67,6 @@ spec:
key: {{ .Values.notify.tokenSecret.key }}
{{- end }}
{{- end }}
volumeMounts:
- name: data
mountPath: /data
livenessProbe:
httpGet:
path: /healthz
@@ -110,25 +78,4 @@ spec:
port: http
initialDelaySeconds: 5
{{- if .Values.backupSidecar.enabled }}
# Idle sidecar. It exists only so k8up has a container with a sqlite3
# module to exec the backupcommand in. Mounted read-write on purpose:
# see the note on the backupcommand annotation above.
- name: backup
image: "{{ .Values.backupSidecar.image.repository }}:{{ .Values.backupSidecar.image.tag }}"
imagePullPolicy: {{ .Values.backupSidecar.image.pullPolicy }}
command: ["sleep", "infinity"]
volumeMounts:
- name: data
mountPath: /data
resources:
requests:
memory: "16Mi"
cpu: "10m"
limits:
memory: "64Mi"
{{- end }}
volumes:
- name: data
persistentVolumeClaim:
claimName: {{ .Release.Name }}-data
-13
View File
@@ -1,13 +0,0 @@
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: {{ .Release.Name }}-data
namespace: {{ .Release.Namespace }}
spec:
storageClassName: {{ .Values.storage.storageClass | quote }}
accessModes:
- ReadWriteOnce
resources:
requests:
storage: {{ .Values.storage.size }}
+14 -14
View File
@@ -11,9 +11,16 @@ image:
tag: "latest"
pullPolicy: IfNotPresent
storage:
size: 1Gi
storageClass: synology-iscsi
# Postgres connection, as a DSN in an existing Secret:
# postgres://user:password@host:5432/terdut?sslmode=require
#
# The chart provisions no database. In this cluster the wrapper chart declares an
# acid.zalan.do postgresql CR and points this at the Secret the operator writes;
# anywhere else, any reachable Postgres will do.
database:
dsnSecret:
name: ""
key: dsn
service:
type: ClusterIP
@@ -88,17 +95,10 @@ notify:
name: ""
key: token
# The server image is FROM scratch — just the binary, with no shell, no sqlite3
# and no python — so a k8up backupcommand cannot run in the app container. This
# idle sidecar shares the data volume and is selected with
# k8up.io/backupcommand-container. Only the stdlib sqlite3 module is used, so any
# python image works.
backupSidecar:
enabled: true
image:
repository: python
tag: "3.13-alpine"
pullPolicy: IfNotPresent
# Backups are no longer this chart's business. The SQLite database lived on a PVC
# beside the app, so it needed a sidecar with a sqlite3 module for k8up to exec a
# dump in; Postgres is backed up where it runs, through a k8up.io/backupcommand
# pg_dump annotation on the database pod itself.
bootstrap:
enabled: true