目录
前言
Memcached的LRU算法分析
Memcached的LRU几种策略
LRU的基本操作和存储的数据结构
策略1 - 惰性删除
策略2 - flush命令
策略3 - 分配Item的时候去检查
策略4 - LRU爬虫
上一章《Memcached源码分析 - Memcached源码分析之增删改查操作(5) 》中,我们讲到了SET命令的操作。当客户端向Memcached服务端SET一条缓存数据的时候,会将生成的Item地址挂到LRU的链表结构上。这一章节,我们主要讲一下Memcached是如何使用LRU算法的。
LRU:是Least Recently Used 近期最少使用算法。
Mecached的LRU的链表操作主要在item.c这个文件上的。其中数组heads和tails分别存储不同的LRU的双向链表的头地址和尾部地址。
每个slabs class都会有自己的一个双向链表结构。链表结构主要通过item结构中的两个指针地址来记录item在链表上左右两边位置的item地址值。
//item的具体结构
typedef struct _stritem {
//记录LRU双向链表下一个item的地址
struct _stritem *next; //下一个结构
//记录LRU双向链表前一个Item的地址
struct _stritem *prev; //前一个结构
//....more code
} item;
static item *heads[LARGEST_ID]; //存储头部地址
static item *tails[LARGEST_ID]; //存储尾部地址
item_link_q方法主要是将一个item添加到LRU链表上面
//从LRU链表上新增一个Item
//LRU链表是一个双向链表结构
static void item_link_q(item *it) { /* item is the new head */
item **head, **tail;
assert(it->slabs_clsid < LARGEST_ID);
assert((it->it_flags & ITEM_SLABBED) == 0);
head = &heads[it->slabs_clsid];
tail = &tails[it->slabs_clsid];
assert(it != *head);
assert((*head && *tail) || (*head == 0 && *tail == 0));
it->prev = 0;
it->next = *head;
if (it->next) it->next->prev = it;
*head = it;
if (*tail == 0) *tail = it;
sizes[it->slabs_clsid]++;
return;
}
item_unlink_q方法主要是将一个item从LRU链表上面解除:
//从LRU链表上解除Item
static void item_unlink_q(item *it) {
item **head, **tail;
assert(it->slabs_clsid < LARGEST_ID);
head = &heads[it->slabs_clsid];
tail = &tails[it->slabs_clsid];
if (*head == it) {
assert(it->prev == 0);
*head = it->next;
}
if (*tail == it) {
assert(it->next == 0);
*tail = it->prev;
}
assert(it->next != it);
assert(it->prev != it);
if (it->next) it->next->prev = it->prev;
if (it->prev) it->prev->next = it->next;
sizes[it->slabs_clsid]--;
return;
}
Memcached的缓存清除策略是惰性的。这个如何来理解?当用户设置了一个缓存数据,缓存有效期为5分钟。当5分钟时间过后,缓存失效,这个时候Memcached并不会自动去检查当前的Item是否过期。当客户端再次来请求这个数据的时候,才会去检查缓存是否失效了,如果失效则会去清除这个数据。
看一下do_item_get这个方法中,判断缓存数据是否失效的代码:
/** wrapper around assoc_find which does the lazy expiration logic */
item *do_item_get(const char *key, const size_t nkey, const uint32_t hv) {
//...code
if (it != NULL) {
//settings.oldest_live主要用来记录flush命令执行的时间
//it->time用来记录item最近set/add/replce等操作的时间(get操作不会改变)
//然后判断it->time是否在执行flush命令之前,如果是执行flush之前,说明该item已经失效
if (settings.oldest_live != 0 && settings.oldest_live <= current_time &&
it->time <= settings.oldest_live) {
//LRU链表和HASHTABLE上解除绑定
do_item_unlink(it, hv);
//删除该Item
do_item_remove(it);
it = NULL; //返回NULL
if (was_found) {
fprintf(stderr, " -nuked by flush");
}
//检查是否过期,主要是检查有效期时间
//如果数据已经过期,则需要清除
} else if (it->exptime != 0 && it->exptime <= current_time) {
//LRU链表和HASHTABLE上解除绑定
do_item_unlink(it, hv);
//删除该Item
do_item_remove(it);
it = NULL;
if (was_found) {
fprintf(stderr, " -nuked by expire");
}
} else {
it->it_flags |= ITEM_FETCHED;
DEBUG_REFCNT(it, '+');
}
}
//...code
}
当用户发送一个flush命令的时候,Memcached会将命令之前的所有的缓存都设置为失效。
Memcached不会主动去清除这些item。主要通过两种方式:
1. do_item_flush_expired方法。
Memcached会在接受到flush命令的时候,将设置全局参数settings.oldest_live =current_time - 1。然后去调用item_flush_expired方法。
因为设置全局参数item_flush_expired到调用缓存锁方法之间会有一定的时间差,有可能这个过程中,会有新的item在操作。
然后Memcached调用do_item_flush_expired方法,去遍历所有的LRU链表。do_item_flush_expired不会将每一个在flush命令前的Item删除,因为这样会非常耗时,而是删除在设置全局变量到加上缓存锁这之间操作的item。这样就能加快flush的速度。
2. 惰性删除方法。
Memcached会在get操作的时候去判断it->time是否小于settings.oldest_live,如果小于,说明这个item就是过期的。通过这种方法,惰性删除大批量的item数据。
/*
* Flushes expired items after a flush_all call
*/
void item_flush_expired() {
mutex_lock(&cache_lock);
do_item_flush_expired();
mutex_unlock(&cache_lock);
}
/* expires items that are more recent than the oldest_live setting. */
void do_item_flush_expired(void) {
int i;
item *iter, *next;
if (settings.oldest_live == 0)
return;
for (i = 0; i < LARGEST_ID; i++) {
/* The LRU is sorted in decreasing time order, and an item's timestamp
* is never newer than its last access time, so we only need to walk
* back until we hit an item older than the oldest_live time.
* The oldest_live checking will auto-expire the remaining items.
*/
for (iter = heads[i]; iter != NULL; iter = next) {
/* iter->time of 0 are magic objects. */
//iter->time 最近一次的访问时间
//这边为何是iter->time >= settings.oldest_live?
//因为在执行do_item_flush_expired方法前,已经上了cache锁,其它worker是不能操作的
//这边过程中,如果遍历每一个Item都去删除,那么这个遍历过程会非常缓慢,会导致客户端一直等待。
//
//Memcached就想出了一个聪明的办法,从设置settings.oldest_live到上锁之间,还是会有其它客户端
//操作item数据,那么Memcache就将这一部分数据先清理(这部分数据非常少量),这样就能加快flush的速度
//而剩余iter->time < settings.oldest_live的那大批量的item,会通过惰性删除的方式,在get请求中去判断处理
if (iter->time != 0 && iter->time >= settings.oldest_live) {
next = iter->next;
if ((iter->it_flags & ITEM_SLABBED) == 0) {
do_item_unlink_nolock(iter, hash(ITEM_key(iter), iter->nkey));
}
} else {
/* We've hit the first old item. Continue to the next queue. */
break;
}
}
}
}
Memcached在分配一个新的Item。(这个流程有点绕,需要看N遍,才能明白)步骤如下:
//创建一个新的Item
item *do_item_alloc(char *key, const size_t nkey, const int flags,
const rel_time_t exptime, const int nbytes,
const uint32_t cur_hv) {
uint8_t nsuffix;
item *it = NULL; //item结构
char suffix[40];
//item_make_header 计算存储数据的总长度
size_t ntotal = item_make_header(nkey + 1, flags, nbytes, suffix, &nsuffix);
if (settings.use_cas) {
ntotal += sizeof(uint64_t);
}
//通过ntotal 查询在哪个slabs_class上面
//Memcached会根据存储数据长度的不同,分为N多个slabs_class
//用户存储数据的时候,根据需要存储数据的长度,就可以查询到需要存储到哪个slabs_class中。
//每个slabs_class都由诺干个slabs组成,slabs每个大小为1M,我们的item结构的数据就会被分配在slabs上
//每个slabs都会根据自己slabs_class存储的数据块的大小,会被分割为诺干个chunk
//
//举个例子:
//如果id=1的slabs_class为存储 最大为224个字节的缓存数据
//当用户的设置的缓存数据总数据长度为200个字节,则这个item结构就会存储到id=1的slabs_class上。
//当第一次或者slabs_class中的slabs不够用的时候,slabs_class就会去分配一个1M的slabs给存储item使用
//因为id=1的slabs_class存储小于224个字节的数据,所以slabs会被分割为诺干个大小为224字节的chunk块
//我们的item结构数据,就会存储在这个chunk块上面
unsigned int id = slabs_clsid(ntotal);
if (id == 0)
return 0;
mutex_lock(&cache_lock);
/* do a quick check if we have any expired items in the tail.. */
int tries = 5;
/* Avoid hangs if a slab has nothing but refcounted stuff in it. */
int tries_lrutail_reflocked = 1000;
int tried_alloc = 0;
item *search;
item *next_it;
void *hold_lock = NULL;
rel_time_t oldest_live = settings.oldest_live;
//这边就可以得到slabs_class上第一个item的地址
//item数据结构通过item->next和item->prev 来记录链表结构
//这边是寻找LRU 链表的尾部地址
search = tails[id];
/* We walk up *only* for locked items. Never searching for expired.
* Waste of CPU for almost all deployments */
//tries = 5 这边只尝试5次循环搜索
//search = tails[id] 搜索从LRU链表 的尾部开始
for (; tries > 0 && search != NULL; tries--, search=next_it) {
/* we might relink search mid-loop, so search->prev isn't reliable */
next_it = search->prev;
if (search->nbytes == 0 && search->nkey == 0 && search->it_flags == 1) {
/* We are a crawler, ignore it. */
tries++;
continue;
}
uint32_t hv = hash(ITEM_key(search), search->nkey);
/* Attempt to hash item lock the "search" item. If locked, no
* other callers can incr the refcount
*/
/* Don't accidentally grab ourselves, or bail if we can't quicklock */
if (hv == cur_hv || (hold_lock = item_trylock(hv)) == NULL)
continue;
/* Now see if the item is refcount locked */
//一般情况下search->refcount为1,如果增加了refcount之后,不等于2,说明item被其它的worker线程锁定
//refcount往上加1,是锁定当前的item,如果不等于2,说明锁定失败
if (refcount_incr(&search->refcount) != 2) {
/* Avoid pathological case with ref'ed items in tail */
do_item_update_nolock(search);
tries_lrutail_reflocked--;
tries++; //try的次数+1
refcount_decr(&search->refcount); //减去1
itemstats[id].lrutail_reflocked++;
/* Old rare bug could cause a refcount leak. We haven't seen
* it in years, but we leave this code in to prevent failures
* just in case */
if (settings.tail_repair_time &&
search->time + settings.tail_repair_time < current_time) {
itemstats[id].tailrepairs++;
search->refcount = 1;
do_item_unlink_nolock(search, hv);
}
if (hold_lock)
item_trylock_unlock(hold_lock);
if (tries_lrutail_reflocked < 1)
break;
continue;
}
/* Expired or flushed */
//这边判断尾部的Item是否失效,如果已经失效了的话,将当前的失效的item分配给最新的缓存
if ((search->exptime != 0 && search->exptime < current_time)
|| (search->time <= oldest_live && oldest_live <= current_time)) {
itemstats[id].reclaimed++;
if ((search->it_flags & ITEM_FETCHED) == 0) {
itemstats[id].expired_unfetched++;
}
it = search;
slabs_adjust_mem_requested(it->slabs_clsid, ITEM_ntotal(it), ntotal);
do_item_unlink_nolock(it, hv);
/* Iniialize the item block: */
it->slabs_clsid = 0;
//slabs_alloc方法是去分配一个新的内存块
} else if ((it = slabs_alloc(ntotal, id)) == NULL) {
tried_alloc = 1;
//如果设置了不允许LRU淘汰,则返回ERROR
if (settings.evict_to_free == 0) {
itemstats[id].outofmemory++;
} else {
//这边设置了LRU淘汰
//如果分配失败,则从LRU链表尾部,淘汰一个item
//如果这个item设置了有效期为0,也会被淘汰
itemstats[id].evicted++;
itemstats[id].evicted_time = current_time - search->time;
if (search->exptime != 0)
itemstats[id].evicted_nonzero++;
if ((search->it_flags & ITEM_FETCHED) == 0) {
itemstats[id].evicted_unfetched++;
}
//这边直接将LRU尾部的ITEM淘汰,并且给了最新的ITEM使用
it = search;
//重新计算一下这个slabclass_t分配出去的内存大小
//直接霸占被淘汰的item就需要重新计算
slabs_adjust_mem_requested(it->slabs_clsid, ITEM_ntotal(it), ntotal);
//从哈希表和lru链表中删除
//it->refcount的值为2,所以item不会被删除,只是HashTable和LRU上的链接关系
do_item_unlink_nolock(it, hv);
/* Initialize the item block: */
it->slabs_clsid = 0;
/* If we've just evicted an item, and the automover is set to
* angry bird mode, attempt to rip memory into this slab class.
* TODO: Move valid object detection into a function, and on a
* "successful" memory pull, look behind and see if the next alloc
* would be an eviction. Then kick off the slab mover before the
* eviction happens.
*/
if (settings.slab_automove == 2)
slabs_reassign(-1, id);
}
}
//解除引用锁定
refcount_decr(&search->refcount);
/* If hash values were equal, we don't grab a second lock */
if (hold_lock)
item_trylock_unlock(hold_lock);
break;
}
/* 如果分配了5次,结果LRU链表尾部的item都是被锁定的,则重新分配一个item */
if (!tried_alloc && (tries == 0 || search == NULL))
it = slabs_alloc(ntotal, id);
if (it == NULL) {
itemstats[id].outofmemory++;
mutex_unlock(&cache_lock);
return NULL;
}
assert(it->slabs_clsid == 0);
assert(it != heads[id]);
/* Item initialization can happen outside of the lock; the item's already
* been removed from the slab LRU.
*/
it->refcount = 1; //引用的次数 又设置为1 /* the caller will have a reference */
mutex_unlock(&cache_lock);
it->next = it->prev = it->h_next = 0;
it->slabs_clsid = id;
DEBUG_REFCNT(it, '*');
it->it_flags = settings.use_cas ? ITEM_CAS : 0;
it->nkey = nkey;
it->nbytes = nbytes;
//这边是内存拷贝,拷贝到item结构地址的内存块上
memcpy(ITEM_key(it), key, nkey);
it->exptime = exptime;
//这边也是内存拷贝
memcpy(ITEM_suffix(it), suffix, (size_t)nsuffix);
it->nsuffix = nsuffix;
return it;
}
//LRU爬虫
static void *item_crawler_thread(void *arg) {
int i;
pthread_mutex_lock(&lru_crawler_lock);
if (settings.verbose > 2)
fprintf(stderr, "Starting LRU crawler background thread\n");
while (do_run_lru_crawler_thread) {
pthread_cond_wait(&lru_crawler_cond, &lru_crawler_lock);
while (crawler_count) {
item *search = NULL;
void *hold_lock = NULL;
for (i = 0; i < LARGEST_ID; i++) {
if (crawlers[i].it_flags != 1) {
continue;
}
pthread_mutex_lock(&cache_lock);
search = crawler_crawl_q((item *)&crawlers[i]);
if (search == NULL ||
(crawlers[i].remaining && --crawlers[i].remaining < 1)) {
if (settings.verbose > 2)
fprintf(stderr, "Nothing left to crawl for %d\n", i);
crawlers[i].it_flags = 0;
crawler_count--;
crawler_unlink_q((item *)&crawlers[i]);
pthread_mutex_unlock(&cache_lock);
continue;
}
uint32_t hv = hash(ITEM_key(search), search->nkey);
/* Attempt to hash item lock the "search" item. If locked, no
* other callers can incr the refcount
*/
if ((hold_lock = item_trylock(hv)) == NULL) {
pthread_mutex_unlock(&cache_lock);
continue;
}
/* Now see if the item is refcount locked */
if (refcount_incr(&search->refcount) != 2) {
refcount_decr(&search->refcount);
if (hold_lock)
item_trylock_unlock(hold_lock);
pthread_mutex_unlock(&cache_lock);
continue;
}
/* Frees the item or decrements the refcount. */
/* Interface for this could improve: do the free/decr here
* instead? */
item_crawler_evaluate(search, hv, i);
if (hold_lock)
item_trylock_unlock(hold_lock);
pthread_mutex_unlock(&cache_lock);
if (settings.lru_crawler_sleep)
usleep(settings.lru_crawler_sleep);
}
}
if (settings.verbose > 2)
fprintf(stderr, "LRU crawler thread sleeping\n");
STATS_LOCK();
stats.lru_crawler_running = false;
STATS_UNLOCK();
}
pthread_mutex_unlock(&lru_crawler_lock);
if (settings.verbose > 2)
fprintf(stderr, "LRU crawler thread stopping\n");
return NULL;
}
int start_item_crawler_thread(void) {
int ret;
if (settings.lru_crawler)
return -1;
pthread_mutex_lock(&lru_crawler_lock);
do_run_lru_crawler_thread = 1;
settings.lru_crawler = true;
if ((ret = pthread_create(&item_crawler_tid, NULL,
item_crawler_thread, NULL)) != 0) {
fprintf(stderr, "Can't create LRU crawler thread: %s\n",
strerror(ret));
pthread_mutex_unlock(&lru_crawler_lock);
return -1;
}
pthread_mutex_unlock(&lru_crawler_lock);
return 0;
}