Update database migrations, remove goose (#2264)

* Add new db migration

* Update migrations
Remove goose

* Add possibility to test direct upgrades

* Try to fix WASM test

* Add checks for specific migrations

* Remove AddMigration
Use WithTransaction
Add Dendrite version to table

* Fix linter issues

* Update tests

* Update comments, outdent if

* Namespace migrations

* Add direct upgrade tests, skipping over one version

* Split migrations

* Update go version in CI

* Fix copy&paste mistake

* Use contexts in migrations

Co-authored-by: kegsay <kegan@matrix.org>
Co-authored-by: Neil Alexander <neilalexander@users.noreply.github.com>
This commit is contained in:
Till 2022-07-25 11:39:22 +02:00 committed by GitHub
parent c7d978274d
commit 081f5e7226
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
58 changed files with 734 additions and 839 deletions

View file

@ -37,6 +37,7 @@ var (
flagBuildConcurrency = flag.Int("build-concurrency", runtime.NumCPU(), "The amount of build concurrency when building images")
flagHead = flag.String("head", "", "Location to a dendrite repository to treat as HEAD instead of Github")
flagDockerHost = flag.String("docker-host", "localhost", "The hostname of the docker client. 'localhost' if running locally, 'host.docker.internal' if running in Docker.")
flagDirect = flag.Bool("direct", false, "If a direct upgrade from the defined FROM version to TO should be done")
alphaNumerics = regexp.MustCompile("[^a-zA-Z0-9]+")
)
@ -229,7 +230,7 @@ func getAndSortVersionsFromGithub(httpClient *http.Client) (semVers []*semver.Ve
return semVers, nil
}
func calculateVersions(cli *http.Client, from, to string) []string {
func calculateVersions(cli *http.Client, from, to string, direct bool) []string {
semvers, err := getAndSortVersionsFromGithub(cli)
if err != nil {
log.Fatalf("failed to collect semvers from github: %s", err)
@ -284,6 +285,9 @@ func calculateVersions(cli *http.Client, from, to string) []string {
if to == HEAD {
versions = append(versions, HEAD)
}
if direct {
versions = []string{versions[0], versions[len(versions)-1]}
}
return versions
}
@ -461,7 +465,7 @@ func main() {
os.Exit(1)
}
cleanup(dockerClient)
versions := calculateVersions(httpClient, *flagFrom, *flagTo)
versions := calculateVersions(httpClient, *flagFrom, *flagTo, *flagDirect)
log.Printf("Testing dendrite versions: %v\n", versions)
branchToImageID := buildDendriteImages(httpClient, dockerClient, *flagTempDir, *flagBuildConcurrency, versions)

View file

@ -1,109 +0,0 @@
## Database migrations
We use [goose](https://github.com/pressly/goose) to handle database migrations. This allows us to execute
both SQL deltas (e.g `ALTER TABLE ...`) as well as manipulate data in the database in Go using Go functions.
To run a migration, the `goose` binary in this directory needs to be built:
```
$ go build ./cmd/goose
```
This binary allows Dendrite databases to be upgraded and downgraded. Sample usage for upgrading the roomserver database:
```
# for sqlite
$ ./goose -dir roomserver/storage/sqlite3/deltas sqlite3 ./roomserver.db up
# for postgres
$ ./goose -dir roomserver/storage/postgres/deltas postgres "user=dendrite dbname=dendrite sslmode=disable" up
```
For a full list of options, including rollbacks, see https://github.com/pressly/goose or use `goose` with no args.
### Rationale
Dendrite creates tables on startup using `CREATE TABLE IF NOT EXISTS`, so you might think that we should also
apply version upgrades on startup as well. This is convenient and doesn't involve an additional binary to run
which complicates upgrades. However, combining the upgrade mechanism and the server binary makes it difficult
to handle rollbacks. Firstly, how do you specify you wish to rollback? We would have to add additional flags
to the main server binary to say "rollback to version X". Secondly, if you roll back the server binary from
version 5 to version 4, the version 4 binary doesn't know how to rollback the database from version 5 to
version 4! For these reasons, we prefer to have a separate "upgrade" binary which is run for database upgrades.
Rather than roll-our-own migration tool, we decided to use [goose](https://github.com/pressly/goose) as it supports
complex migrations in Go code in addition to just executing SQL deltas. Other alternatives like
`github.com/golang-migrate/migrate` [do not support](https://github.com/golang-migrate/migrate/issues/15) these
kinds of complex migrations.
### Adding new deltas
You can add `.sql` or `.go` files manually or you can use goose to create them for you.
If you only want to add a SQL delta then run:
```
$ ./goose -dir serverkeyapi/storage/sqlite3/deltas sqlite3 ./foo.db create new_col sql
2020/09/09 14:37:43 Created new file: serverkeyapi/storage/sqlite3/deltas/20200909143743_new_col.sql
```
In this case, the version number is `20200909143743`. The important thing is that it is always increasing.
Then add up/downgrade SQL commands to the created file which looks like:
```sql
-- +goose Up
-- +goose StatementBegin
SELECT 'up SQL query';
-- +goose StatementEnd
-- +goose Down
-- +goose StatementBegin
SELECT 'down SQL query';
-- +goose StatementEnd
```
You __must__ keep the `+goose` annotations. You'll need to repeat this process for Postgres.
For complex Go migrations:
```
$ ./goose -dir serverkeyapi/storage/sqlite3/deltas sqlite3 ./foo.db create complex_update go
2020/09/09 14:40:38 Created new file: serverkeyapi/storage/sqlite3/deltas/20200909144038_complex_update.go
```
Then modify the created `.go` file which looks like:
```go
package migrations
import (
"database/sql"
"fmt"
"github.com/pressly/goose"
)
func init() {
goose.AddMigration(upComplexUpdate, downComplexUpdate)
}
func upComplexUpdate(tx *sql.Tx) error {
// This code is executed when the migration is applied.
return nil
}
func downComplexUpdate(tx *sql.Tx) error {
// This code is executed when the migration is rolled back.
return nil
}
```
You __must__ import the package in `/cmd/goose/main.go` so `func init()` gets called.
#### Database limitations
- SQLite3 does NOT support `ALTER TABLE table_name DROP COLUMN` - you would have to rename the column or drop the table
entirely and recreate it. ([example](https://github.com/matrix-org/dendrite/blob/master/userapi/storage/accounts/sqlite3/deltas/20200929203058_is_active.sql))
More information: [sqlite.org](https://www.sqlite.org/lang_altertable.html)

View file

@ -1,154 +0,0 @@
// This is custom goose binary
package main
import (
"flag"
"fmt"
"log"
"os"
"github.com/pressly/goose"
pgusers "github.com/matrix-org/dendrite/userapi/storage/postgres/deltas"
slusers "github.com/matrix-org/dendrite/userapi/storage/sqlite3/deltas"
_ "github.com/lib/pq"
_ "github.com/mattn/go-sqlite3"
)
const (
AppService = "appservice"
FederationSender = "federationapi"
KeyServer = "keyserver"
MediaAPI = "mediaapi"
RoomServer = "roomserver"
SigningKeyServer = "signingkeyserver"
SyncAPI = "syncapi"
UserAPI = "userapi"
)
var (
dir = flags.String("dir", "", "directory with migration files")
flags = flag.NewFlagSet("goose", flag.ExitOnError)
component = flags.String("component", "", "dendrite component name")
knownDBs = []string{
AppService, FederationSender, KeyServer, MediaAPI, RoomServer, SigningKeyServer, SyncAPI, UserAPI,
}
)
// nolint: gocyclo
func main() {
err := flags.Parse(os.Args[1:])
if err != nil {
panic(err.Error())
}
args := flags.Args()
if len(args) < 3 {
fmt.Println(
`Usage: goose [OPTIONS] DRIVER DBSTRING COMMAND
Drivers:
postgres
sqlite3
Examples:
goose -component roomserver sqlite3 ./roomserver.db status
goose -component roomserver sqlite3 ./roomserver.db up
goose -component roomserver postgres "user=dendrite dbname=dendrite sslmode=disable" status
Options:
-component string
Dendrite component name e.g roomserver, signingkeyserver, clientapi, syncapi
-table string
migrations table name (default "goose_db_version")
-h print help
-v enable verbose mode
-dir string
directory with migration files, only relevant when creating new migrations.
-version
print version
Commands:
up Migrate the DB to the most recent version available
up-by-one Migrate the DB up by 1
up-to VERSION Migrate the DB to a specific VERSION
down Roll back the version by 1
down-to VERSION Roll back to a specific VERSION
redo Re-run the latest migration
reset Roll back all migrations
status Dump the migration status for the current DB
version Print the current version of the database
create NAME [sql|go] Creates new migration file with the current timestamp
fix Apply sequential ordering to migrations`,
)
return
}
engine := args[0]
if engine != "sqlite3" && engine != "postgres" {
fmt.Println("engine must be one of 'sqlite3' or 'postgres'")
return
}
knownComponent := false
for _, c := range knownDBs {
if c == *component {
knownComponent = true
break
}
}
if !knownComponent {
fmt.Printf("component must be one of %v\n", knownDBs)
return
}
if engine == "sqlite3" {
loadSQLiteDeltas(*component)
} else {
loadPostgresDeltas(*component)
}
dbstring, command := args[1], args[2]
db, err := goose.OpenDBWithDriver(engine, dbstring)
if err != nil {
log.Fatalf("goose: failed to open DB: %v\n", err)
}
defer func() {
if err := db.Close(); err != nil {
log.Fatalf("goose: failed to close DB: %v\n", err)
}
}()
arguments := []string{}
if len(args) > 3 {
arguments = append(arguments, args[3:]...)
}
// goose demands a directory even though we don't use it for upgrades
d := *dir
if d == "" {
d = os.TempDir()
}
if err := goose.Run(command, db, d, arguments...); err != nil {
log.Fatalf("goose %v: %v", command, err)
}
}
func loadSQLiteDeltas(component string) {
switch component {
case UserAPI:
slusers.LoadFromGoose()
}
}
func loadPostgresDeltas(component string) {
switch component {
case UserAPI:
pgusers.LoadFromGoose()
}
}