[performance] update remaining worker pools to use queues (#2865)

* start replacing client + federator + media workers with new worker + queue types

* refactor federatingDB.Delete(), drop queued messages when deleting account / status

* move all queue purging to the processor workers

* undo toolchain updates

* code comments, ensure dereferencer worker pool gets started

* update gruf libraries in readme

* start the job scheduler separately to the worker pools

* reshuffle ordering or server.go + remove duplicate worker start / stop

* update go-list version

* fix vendoring

* move queue invalidation to before wipeing / deletion, to ensure queued work not dropped

* add logging to worker processing functions in testrig, don't start workers in unexpected places

* update go-structr to add (+then rely on) QueueCtx{} type

* ensure more worker pools get started properly in tests

* fix remaining broken tests relying on worker queue logic

* fix account test suite queue popping logic, ensure noop workers do not pull from queue

* move back accidentally shuffled account deletion order

* ensure error (non nil!!) gets passed in refactored federatingDB{}.Delete()

* silently drop deletes from accounts not permitted to

* don't warn log on forwarded deletes

* make if else clauses easier to parse

* use getFederatorMsg()

* improved code comment

* improved code comment re: requesting account delete checks

* remove boolean result from worker start / stop since false = already running or already stopped

* remove optional passed-in http.client

* remove worker starting from the admin CLI commands (we don't need to handle side-effects)

* update prune cli to start scheduler but not all of the workers

* fix rebase issues

* remove redundant return statements

* i'm sorry sir linter
This commit is contained in:
kim 2024-04-26 13:50:46 +01:00 committed by GitHub
commit c9c0773f2c
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
79 changed files with 1892 additions and 834 deletions

View file

@ -0,0 +1,142 @@
// GoToSocial
// Copyright (C) GoToSocial Authors admin@gotosocial.org
// SPDX-License-Identifier: AGPL-3.0-or-later
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Affero General Public License for more details.
//
// You should have received a copy of the GNU Affero General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
package workers
import (
"context"
"codeberg.org/gruf/go-runners"
"github.com/superseriousbusiness/gotosocial/internal/log"
"github.com/superseriousbusiness/gotosocial/internal/queue"
"github.com/superseriousbusiness/gotosocial/internal/util"
)
// FnWorkerPool wraps multiple FnWorker{}s in
// a singular struct for easy multi start / stop.
type FnWorkerPool struct {
// Queue is embedded queue.SimpleQueue{}
// passed to each of the pool Worker{}s.
Queue queue.SimpleQueue[func(context.Context)]
// internal fields.
workers []*FnWorker
}
// Start will attempt to start 'n' FnWorker{}s.
func (p *FnWorkerPool) Start(n int) {
// Check whether workers are
// set (is already running).
ok := (len(p.workers) > 0)
if ok {
return
}
// Allocate new workers slice.
p.workers = make([]*FnWorker, n)
for i := range p.workers {
// Allocate new FnWorker{}.
p.workers[i] = new(FnWorker)
p.workers[i].Queue = &p.Queue
// Attempt to start worker.
// Return bool not useful
// here, as true = started,
// false = already running.
_ = p.workers[i].Start()
}
}
// Stop will attempt to stop contained FnWorker{}s.
func (p *FnWorkerPool) Stop() {
// Check whether workers are
// set (is currently running).
ok := (len(p.workers) == 0)
if ok {
return
}
// Stop all running workers.
for i := range p.workers {
// return bool not useful
// here, as true = stopped,
// false = never running.
_ = p.workers[i].Stop()
}
// Unset workers slice.
p.workers = p.workers[:0]
}
// FnWorker wraps a queue.SimpleQueue{} which
// it feeds from to provide it with function
// tasks to execute. It does so in a single
// goroutine with state management utilities.
type FnWorker struct {
// Queue is the fn queue that FnWorker
// will feed from for upcoming tasks.
Queue *queue.SimpleQueue[func(context.Context)]
// internal fields.
service runners.Service
}
// Start will attempt to start the Worker{}.
func (w *FnWorker) Start() bool {
return w.service.GoRun(w.run)
}
// Stop will attempt to stop the Worker{}.
func (w *FnWorker) Stop() bool {
return w.service.Stop()
}
// run wraps process to restart on any panic.
func (w *FnWorker) run(ctx context.Context) {
if w.Queue == nil {
panic("not yet initialized")
}
log.Infof(ctx, "%p: starting worker", w)
defer log.Infof(ctx, "%p: stopped worker", w)
util.Must(func() { w.process(ctx) })
}
// process is the main delivery worker processing routine.
func (w *FnWorker) process(ctx context.Context) {
if w.Queue == nil {
// we perform this check here just
// to ensure the compiler knows these
// variables aren't nil in the loop,
// even if already checked by caller.
panic("not yet initialized")
}
for {
// Block until pop next func.
fn, ok := w.Queue.PopCtx(ctx)
if !ok {
return
}
// run!
fn(ctx)
}
}

View file

@ -0,0 +1,157 @@
// GoToSocial
// Copyright (C) GoToSocial Authors admin@gotosocial.org
// SPDX-License-Identifier: AGPL-3.0-or-later
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Affero General Public License for more details.
//
// You should have received a copy of the GNU Affero General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
package workers
import (
"context"
"codeberg.org/gruf/go-runners"
"codeberg.org/gruf/go-structr"
"github.com/superseriousbusiness/gotosocial/internal/log"
"github.com/superseriousbusiness/gotosocial/internal/queue"
"github.com/superseriousbusiness/gotosocial/internal/util"
)
// MsgWorkerPool wraps multiple MsgWorker{}s in
// a singular struct for easy multi start / stop.
type MsgWorkerPool[Msg any] struct {
// Process handles queued message types.
Process func(context.Context, Msg) error
// Queue is embedded queue.StructQueue{}
// passed to each of the pool Worker{}s.
Queue queue.StructQueue[Msg]
// internal fields.
workers []*MsgWorker[Msg]
}
// Init will initialize the worker pool queue with given struct indices.
func (p *MsgWorkerPool[T]) Init(indices []structr.IndexConfig) {
p.Queue.Init(structr.QueueConfig[T]{Indices: indices})
}
// Start will attempt to start 'n' Worker{}s.
func (p *MsgWorkerPool[T]) Start(n int) {
// Check whether workers are
// set (is already running).
ok := (len(p.workers) > 0)
if ok {
return
}
// Allocate new msg workers slice.
p.workers = make([]*MsgWorker[T], n)
for i := range p.workers {
// Allocate new MsgWorker[T]{}.
p.workers[i] = new(MsgWorker[T])
p.workers[i].Process = p.Process
p.workers[i].Queue = &p.Queue
// Attempt to start worker.
// Return bool not useful
// here, as true = started,
// false = already running.
_ = p.workers[i].Start()
}
}
// Stop will attempt to stop contained Worker{}s.
func (p *MsgWorkerPool[T]) Stop() {
// Check whether workers are
// set (is currently running).
ok := (len(p.workers) == 0)
if ok {
return
}
// Stop all running workers.
for i := range p.workers {
// return bool not useful
// here, as true = stopped,
// false = never running.
_ = p.workers[i].Stop()
}
// Unset workers slice.
p.workers = p.workers[:0]
}
// MsgWorker wraps a processing function to
// feed from a queue.StructQueue{} for messages
// to process. It does so in a single goroutine
// with state management utilities.
type MsgWorker[Msg any] struct {
// Process handles queued message types.
Process func(context.Context, Msg) error
// Queue is the Delivery{} message queue
// that delivery worker will feed from.
Queue *queue.StructQueue[Msg]
// internal fields.
service runners.Service
}
// Start will attempt to start the Worker{}.
func (w *MsgWorker[T]) Start() bool {
return w.service.GoRun(w.run)
}
// Stop will attempt to stop the Worker{}.
func (w *MsgWorker[T]) Stop() bool {
return w.service.Stop()
}
// run wraps process to restart on any panic.
func (w *MsgWorker[T]) run(ctx context.Context) {
if w.Process == nil || w.Queue == nil {
panic("not yet initialized")
}
log.Infof(ctx, "%p: starting worker", w)
defer log.Infof(ctx, "%p: stopped worker", w)
util.Must(func() { w.process(ctx) })
}
// process is the main delivery worker processing routine.
func (w *MsgWorker[T]) process(ctx context.Context) {
if w.Process == nil || w.Queue == nil {
// we perform this check here just
// to ensure the compiler knows these
// variables aren't nil in the loop,
// even if already checked by caller.
panic("not yet initialized")
}
for {
// Block until pop next message.
msg, ok := w.Queue.PopCtx(ctx)
if !ok {
return
}
// Attempt to process popped message type.
if err := w.Process(ctx, msg); err != nil {
log.Errorf(ctx, "%p: error processing: %v", w, err)
}
}
}

View file

@ -18,11 +18,8 @@
package workers
import (
"context"
"log"
"runtime"
"codeberg.org/gruf/go-runners"
"github.com/superseriousbusiness/gotosocial/internal/config"
"github.com/superseriousbusiness/gotosocial/internal/messages"
"github.com/superseriousbusiness/gotosocial/internal/scheduler"
@ -39,77 +36,49 @@ type Workers struct {
// indexed queue of Delivery{} objects.
Delivery delivery.WorkerPool
// ClientAPI provides a worker pool that handles both
// incoming client actions, and our own side-effects.
ClientAPI runners.WorkerPool
// Client provides a worker pool that handles
// incoming processing jobs from the client API.
Client MsgWorkerPool[*messages.FromClientAPI]
// Federator provides a worker pool that handles both
// incoming federated actions, and our own side-effects.
Federator runners.WorkerPool
// Federator provides a worker pool that handles
// incoming processing jobs from the fedi API.
Federator MsgWorkerPool[*messages.FromFediAPI]
// Enqueue functions for clientAPI / federator worker pools,
// these are pointers to Processor{}.Enqueue___() msg functions.
// This prevents dependency cycling as Processor depends on Workers.
EnqueueClientAPI func(context.Context, ...messages.FromClientAPI)
EnqueueFediAPI func(context.Context, ...messages.FromFediAPI)
// Dereference provides a worker pool
// for asynchronous dereferencer jobs.
Dereference FnWorkerPool
// Blocking processing functions for clientAPI / federator.
// These are pointers to Processor{}.Process___() msg functions.
// This prevents dependency cycling as Processor depends on Workers.
//
// Rather than queueing messages for asynchronous processing, these
// functions will process immediately and in a blocking manner, and
// will not use up a worker slot.
//
// As such, you should only call them in special cases where something
// synchronous needs to happen before you can do something else.
ProcessFromClientAPI func(context.Context, messages.FromClientAPI) error
ProcessFromFediAPI func(context.Context, messages.FromFediAPI) error
// Media manager worker pools.
Media runners.WorkerPool
// Media provides a worker pool for
// asynchronous media processing jobs.
Media FnWorkerPool
// prevent pass-by-value.
_ nocopy
}
// Start will start all of the contained
// worker pools (and global scheduler).
// StartScheduler starts the job scheduler.
func (w *Workers) StartScheduler() {
_ = w.Scheduler.Start() // false = already running
}
// Start will start contained worker pools.
func (w *Workers) Start() {
// Get currently set GOMAXPROCS.
maxprocs := runtime.GOMAXPROCS(0)
tryUntil("starting scheduler", 5, w.Scheduler.Start)
tryUntil("start delivery workerpool", 5, func() bool {
n := config.GetAdvancedSenderMultiplier()
if n < 1 {
// clamp min senders to 1.
return w.Delivery.Start(1)
}
return w.Delivery.Start(n * maxprocs)
})
tryUntil("starting client API workerpool", 5, func() bool {
return w.ClientAPI.Start(4*maxprocs, 400*maxprocs)
})
tryUntil("starting federator workerpool", 5, func() bool {
return w.Federator.Start(4*maxprocs, 400*maxprocs)
})
tryUntil("starting media workerpool", 5, func() bool {
return w.Media.Start(8*maxprocs, 80*maxprocs)
})
w.Delivery.Start(deliveryWorkers(maxprocs))
w.Client.Start(4 * maxprocs)
w.Federator.Start(4 * maxprocs)
w.Dereference.Start(4 * maxprocs)
w.Media.Start(8 * maxprocs)
}
// Stop will stop all of the contained worker pools (and global scheduler).
func (w *Workers) Stop() {
tryUntil("stopping scheduler", 5, w.Scheduler.Stop)
tryUntil("stopping delivery workerpool", 5, w.Delivery.Stop)
tryUntil("stopping client API workerpool", 5, w.ClientAPI.Stop)
tryUntil("stopping federator workerpool", 5, w.Federator.Stop)
tryUntil("stopping media workerpool", 5, w.Media.Stop)
_ = w.Scheduler.Stop() // false = not running
w.Delivery.Stop()
w.Client.Stop()
w.Federator.Stop()
w.Dereference.Stop()
w.Media.Stop()
}
// nocopy when embedded will signal linter to
@ -120,12 +89,11 @@ func (*nocopy) Lock() {}
func (*nocopy) Unlock() {}
// tryUntil will attempt to call 'do' for 'count' attempts, before panicking with 'msg'.
func tryUntil(msg string, count int, do func() bool) {
for i := 0; i < count; i++ {
if do() {
return
}
func deliveryWorkers(maxprocs int) int {
n := config.GetAdvancedSenderMultiplier()
if n < 1 {
// clamp to 1
return 1
}
log.Panicf("failed %s after %d tries", msg, count)
return n * maxprocs
}