fix(permissions): chunked synchronization of role permission events (#9403)

# Which Problems Are Solved

Setup fails to push all role permission events when running Zitadel with
CockroachDB. `TransactionRetryError`s were visible in logs which finally
times out the setup job with `timeout: context deadline exceeded`

# How the Problems Are Solved

As suggested in the [Cockroach documentation](timeout: context deadline
exceeded), _"break down larger transactions"_. The commands to be pushed
for the role permissions are chunked in 50 events per push. This
chunking is only done with CockroachDB.

# Additional Changes

- gci run fixed some unrelated imports
- access to `command.Commands` for the setup job, so we can reuse the
sync logic.

# Additional Context

Closes #9293

---------

Co-authored-by: Silvan <27845747+adlerhurst@users.noreply.github.com>
This commit is contained in:
Tim Möhlmann
2025-02-26 18:06:50 +02:00
committed by GitHub
parent 77499ce603
commit e670b9126c
13 changed files with 461 additions and 169 deletions

View File

@@ -2,29 +2,22 @@ package setup
import (
"context"
"database/sql"
_ "embed"
"fmt"
"strings"
"github.com/zitadel/logging"
"github.com/zitadel/zitadel/internal/api/authz"
"github.com/zitadel/zitadel/internal/database"
"github.com/zitadel/zitadel/internal/command"
"github.com/zitadel/zitadel/internal/eventstore"
"github.com/zitadel/zitadel/internal/repository/instance"
"github.com/zitadel/zitadel/internal/repository/permission"
)
var (
//go:embed sync_role_permissions.sql
getRolePermissionOperationsQuery string
)
// SyncRolePermissions is a repeatable step which synchronizes the InternalAuthZ
// RolePermissionMappings from the configuration to the database.
// This is needed until role permissions are manageable over the API.
type SyncRolePermissions struct {
commands *command.Commands
eventstore *eventstore.Eventstore
rolePermissionMappings []authz.RoleMapping
}
@@ -38,18 +31,11 @@ func (mig *SyncRolePermissions) Execute(ctx context.Context, _ eventstore.Event)
func (mig *SyncRolePermissions) executeSystem(ctx context.Context) error {
logging.WithFields("migration", mig.String()).Info("prepare system role permission sync events")
target := rolePermissionMappingsToDatabaseMap(mig.rolePermissionMappings, true)
cmds, err := mig.synchronizeCommands(ctx, "SYSTEM", target)
details, err := mig.commands.SynchronizeRolePermission(ctx, "SYSTEM", mig.rolePermissionMappings)
if err != nil {
return err
}
events, err := mig.eventstore.Push(ctx, cmds...)
if err != nil {
return err
}
logging.WithFields("migration", mig.String(), "pushed_events", len(events)).Info("pushed system role permission sync events")
logging.WithFields("migration", mig.String(), "sequence", details.Sequence).Info("pushed system role permission sync events")
return nil
}
@@ -70,51 +56,17 @@ func (mig *SyncRolePermissions) executeInstances(ctx context.Context) error {
if err != nil {
return err
}
target := rolePermissionMappingsToDatabaseMap(mig.rolePermissionMappings, false)
for i, instanceID := range instances {
logging.WithFields("instance_id", instanceID, "migration", mig.String(), "progress", fmt.Sprintf("%d/%d", i+1, len(instances))).Info("prepare instance role permission sync events")
cmds, err := mig.synchronizeCommands(ctx, instanceID, target)
details, err := mig.commands.SynchronizeRolePermission(ctx, instanceID, mig.rolePermissionMappings)
if err != nil {
return err
}
events, err := mig.eventstore.Push(ctx, cmds...)
if err != nil {
return err
}
logging.WithFields("instance_id", instanceID, "migration", mig.String(), "pushed_events", len(events)).Info("pushed instance role permission sync events")
logging.WithFields("instance_id", instanceID, "migration", mig.String(), "sequence", details.Sequence).Info("pushed instance role permission sync events")
}
return nil
}
// synchronizeCommands checks the current state of role permissions in the eventstore for the aggregate.
// It returns the commands required to reach the desired state passed in target.
// For system level permissions aggregateID must be set to `SYSTEM`,
// else it is the instance ID.
func (mig *SyncRolePermissions) synchronizeCommands(ctx context.Context, aggregateID string, target database.Map[[]string]) (cmds []eventstore.Command, err error) {
aggregate := permission.NewAggregate(aggregateID)
err = mig.eventstore.Client().QueryContext(ctx, func(rows *sql.Rows) error {
for rows.Next() {
var operation, role, perm string
if err := rows.Scan(&operation, &role, &perm); err != nil {
return err
}
logging.WithFields("aggregate_id", aggregateID, "migration", mig.String(), "operation", operation, "role", role, "permission", perm).Debug("sync role permission")
switch operation {
case "add":
cmds = append(cmds, permission.NewAddedEvent(ctx, aggregate, role, perm))
case "remove":
cmds = append(cmds, permission.NewRemovedEvent(ctx, aggregate, role, perm))
}
}
return rows.Close()
}, getRolePermissionOperationsQuery, aggregateID, target)
if err != nil {
return nil, err
}
return cmds, err
}
func (*SyncRolePermissions) String() string {
return "repeatable_sync_role_permissions"
}
@@ -122,13 +74,3 @@ func (*SyncRolePermissions) String() string {
func (*SyncRolePermissions) Check(lastRun map[string]interface{}) bool {
return true
}
func rolePermissionMappingsToDatabaseMap(mappings []authz.RoleMapping, system bool) database.Map[[]string] {
out := make(database.Map[[]string], len(mappings))
for _, m := range mappings {
if system == strings.HasPrefix(m.Role, "SYSTEM") {
out[m.Role] = m.Permissions
}
}
return out
}