Add LRU mem cache implementation (#16226)

The current default memory cache implementation is unbounded in size and number of
objects cached. This is hardly ideal.

This PR proposes creating a TwoQueue LRU cache as the underlying cache for Gitea.
The cache is limited by the number of objects stored in the cache (rather than size)
for simplicity. The default number of objects is 50000 - which is perhaps too small
as most of our objects cached are going to be much less than 1kB.

It may be worth considering using a different LRU implementation that actively limits
sizes or avoids GC - however, this is just a beginning implementation.

Signed-off-by: Andrew Thornton <art27@cantab.net>
This commit is contained in:
zeripath 2021-07-10 22:54:15 +01:00 committed by GitHub
parent 07284792d4
commit 2f725cbc9e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 219 additions and 6 deletions

View File

@ -1471,7 +1471,7 @@ PATH =
;; if the cache enabled
;ENABLED = true
;;
;; Either "memory", "redis", or "memcache", default is "memory"
;; Either "memory", "redis", "memcache", or "twoqueue". default is "memory"
;ADAPTER = memory
;;
;; For "memory" only, GC interval in seconds, default is 60
@ -1480,6 +1480,7 @@ PATH =
;; For "redis" and "memcache", connection host address
;; redis: network=tcp,addr=:6379,password=macaron,db=0,pool_size=100,idle_timeout=180
;; memcache: `127.0.0.1:11211`
;; twoqueue: `{"size":50000,"recent_ratio":0.25,"ghost_ratio":0.5}` or `50000`
;HOST =
;;
;; Time to keep items in cache if not used, default is 16 hours.

View File

@ -590,11 +590,12 @@ Define allowed algorithms and their minimum key length (use -1 to disable a type
## Cache (`cache`)
- `ENABLED`: **true**: Enable the cache.
- `ADAPTER`: **memory**: Cache engine adapter, either `memory`, `redis`, or `memcache`.
- `INTERVAL`: **60**: Garbage Collection interval (sec), for memory cache only.
- `HOST`: **\<empty\>**: Connection string for `redis` and `memcache`.
- `ADAPTER`: **memory**: Cache engine adapter, either `memory`, `redis`, `twoqueue` or `memcache`. (`twoqueue` represents a size limited LRU cache.)
- `INTERVAL`: **60**: Garbage Collection interval (sec), for memory and twoqueue cache only.
- `HOST`: **\<empty\>**: Connection string for `redis` and `memcache`. For `twoqueue` sets configuration for the queue.
- Redis: `redis://:macaron@127.0.0.1:6379/0?pool_size=100&idle_timeout=180s`
- Memcache: `127.0.0.1:9090;127.0.0.1:9091`
- TwoQueue LRU cache: `{"size":50000,"recent_ratio":0.25,"ghost_ratio":0.5}` or `50000` representing the maximum number of objects stored in the cache.
- `ITEM_TTL`: **16h**: Time to keep items in cache if not used, Setting it to 0 disables caching.
## Cache - LastCommitCache settings (`cache.last_commit`)

204
modules/cache/cache_twoqueue.go vendored Normal file
View File

@ -0,0 +1,204 @@
// Copyright 2021 The Gitea Authors. All rights reserved.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package cache
import (
"strconv"
"sync"
"time"
mc "gitea.com/go-chi/cache"
lru "github.com/hashicorp/golang-lru"
jsoniter "github.com/json-iterator/go"
)
// TwoQueueCache represents a LRU 2Q cache adapter implementation
type TwoQueueCache struct {
lock sync.Mutex
cache *lru.TwoQueueCache
interval int
}
// TwoQueueCacheConfig describes the configuration for TwoQueueCache
type TwoQueueCacheConfig struct {
Size int `ini:"SIZE" json:"size"`
RecentRatio float64 `ini:"RECENT_RATIO" json:"recent_ratio"`
GhostRatio float64 `ini:"GHOST_RATIO" json:"ghost_ratio"`
}
// MemoryItem represents a memory cache item.
type MemoryItem struct {
Val interface{}
Created int64
Timeout int64
}
func (item *MemoryItem) hasExpired() bool {
return item.Timeout > 0 &&
(time.Now().Unix()-item.Created) >= item.Timeout
}
var _ mc.Cache = &TwoQueueCache{}
// Put puts value into cache with key and expire time.
func (c *TwoQueueCache) Put(key string, val interface{}, timeout int64) error {
item := &MemoryItem{
Val: val,
Created: time.Now().Unix(),
Timeout: timeout,
}
c.lock.Lock()
defer c.lock.Unlock()
c.cache.Add(key, item)
return nil
}
// Get gets cached value by given key.
func (c *TwoQueueCache) Get(key string) interface{} {
c.lock.Lock()
defer c.lock.Unlock()
cached, ok := c.cache.Get(key)
if !ok {
return nil
}
item, ok := cached.(*MemoryItem)
if !ok || item.hasExpired() {
c.cache.Remove(key)
return nil
}
return item.Val
}
// Delete deletes cached value by given key.
func (c *TwoQueueCache) Delete(key string) error {
c.lock.Lock()
defer c.lock.Unlock()
c.cache.Remove(key)
return nil
}
// Incr increases cached int-type value by given key as a counter.
func (c *TwoQueueCache) Incr(key string) error {
c.lock.Lock()
defer c.lock.Unlock()
cached, ok := c.cache.Get(key)
if !ok {
return nil
}
item, ok := cached.(*MemoryItem)
if !ok || item.hasExpired() {
c.cache.Remove(key)
return nil
}
var err error
item.Val, err = mc.Incr(item.Val)
return err
}
// Decr decreases cached int-type value by given key as a counter.
func (c *TwoQueueCache) Decr(key string) error {
c.lock.Lock()
defer c.lock.Unlock()
cached, ok := c.cache.Get(key)
if !ok {
return nil
}
item, ok := cached.(*MemoryItem)
if !ok || item.hasExpired() {
c.cache.Remove(key)
return nil
}
var err error
item.Val, err = mc.Decr(item.Val)
return err
}
// IsExist returns true if cached value exists.
func (c *TwoQueueCache) IsExist(key string) bool {
c.lock.Lock()
defer c.lock.Unlock()
cached, ok := c.cache.Peek(key)
if !ok {
return false
}
item, ok := cached.(*MemoryItem)
if !ok || item.hasExpired() {
c.cache.Remove(key)
return false
}
return true
}
// Flush deletes all cached data.
func (c *TwoQueueCache) Flush() error {
c.lock.Lock()
defer c.lock.Unlock()
c.cache.Purge()
return nil
}
func (c *TwoQueueCache) checkAndInvalidate(key interface{}) {
c.lock.Lock()
defer c.lock.Unlock()
cached, ok := c.cache.Peek(key)
if !ok {
return
}
item, ok := cached.(*MemoryItem)
if !ok || item.hasExpired() {
c.cache.Remove(item)
}
}
func (c *TwoQueueCache) startGC() {
if c.interval < 0 {
return
}
for _, key := range c.cache.Keys() {
c.checkAndInvalidate(key)
}
time.AfterFunc(time.Duration(c.interval)*time.Second, c.startGC)
}
// StartAndGC starts GC routine based on config string settings.
func (c *TwoQueueCache) StartAndGC(opts mc.Options) error {
var err error
size := 50000
if opts.AdapterConfig != "" {
size, err = strconv.Atoi(opts.AdapterConfig)
}
if err != nil {
json := jsoniter.ConfigCompatibleWithStandardLibrary
if !json.Valid([]byte(opts.AdapterConfig)) {
return err
}
cfg := &TwoQueueCacheConfig{
Size: 50000,
RecentRatio: lru.Default2QRecentRatio,
GhostRatio: lru.Default2QGhostEntries,
}
_ = json.Unmarshal([]byte(opts.AdapterConfig), cfg)
c.cache, err = lru.New2QParams(cfg.Size, cfg.RecentRatio, cfg.GhostRatio)
} else {
c.cache, err = lru.New2Q(size)
}
c.interval = opts.Interval
if c.interval > 0 {
go c.startGC()
}
return err
}
func init() {
mc.Register("twoqueue", &TwoQueueCache{})
}

View File

@ -58,11 +58,16 @@ func newCacheService() {
log.Fatal("Failed to map Cache settings: %v", err)
}
CacheService.Adapter = sec.Key("ADAPTER").In("memory", []string{"memory", "redis", "memcache"})
CacheService.Adapter = sec.Key("ADAPTER").In("memory", []string{"memory", "redis", "memcache", "twoqueue"})
switch CacheService.Adapter {
case "memory":
case "redis", "memcache":
CacheService.Conn = strings.Trim(sec.Key("HOST").String(), "\" ")
case "twoqueue":
CacheService.Conn = strings.TrimSpace(sec.Key("HOST").String())
if CacheService.Conn == "" {
CacheService.Conn = "50000"
}
case "": // disable cache
CacheService.Enabled = false
default:

View File

@ -52,7 +52,9 @@ func NewServices() {
log.Fatal("repository init failed: %v", err)
}
mailer.NewContext()
_ = cache.NewContext()
if err := cache.NewContext(); err != nil {
log.Fatal("Unable to start cache service: %v", err)
}
notification.NewContext()
if err := archiver.Init(); err != nil {
log.Fatal("archiver init failed: %v", err)