#include <float.h>
#include <math.h>
#include <pthread.h>
+#include <sys/resource.h>
/* Our shared "common" objects */
{"setex",setexCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,0,0,0},
{"append",appendCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1},
{"substr",substrCommand,4,REDIS_CMD_INLINE,NULL,1,1,1},
+ {"strlen",strlenCommand,2,REDIS_CMD_INLINE,NULL,1,1,1},
{"del",delCommand,-2,REDIS_CMD_INLINE,NULL,0,0,0},
{"exists",existsCommand,2,REDIS_CMD_INLINE,NULL,1,1,1},
{"incr",incrCommand,2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,NULL,1,1,1},
{"info",infoCommand,1,REDIS_CMD_INLINE,NULL,0,0,0},
{"monitor",monitorCommand,1,REDIS_CMD_INLINE,NULL,0,0,0},
{"ttl",ttlCommand,2,REDIS_CMD_INLINE,NULL,1,1,1},
+ {"persist",persistCommand,2,REDIS_CMD_INLINE,NULL,1,1,1},
{"slaveof",slaveofCommand,3,REDIS_CMD_INLINE,NULL,0,0,0},
{"debug",debugCommand,-2,REDIS_CMD_INLINE,NULL,0,0,0},
{"config",configCommand,-2,REDIS_CMD_BULK,NULL,0,0,0},
NULL, /* val dup */
dictEncObjKeyCompare, /* key compare */
dictRedisObjectDestructor, /* key destructor */
- dictVanillaFree /* val destructor of malloc(sizeof(double)) */
+ NULL /* val destructor */
};
/* Db->dict, keys are sds strings, vals are Redis objects. */
/* ======================= Cron: called every 100 ms ======================== */
+/* Try to expire a few timed out keys. The algorithm used is adaptive and
+ * will use few CPU cycles if there are few expiring keys, otherwise
+ * it will get more aggressive to avoid that too much memory is used by
+ * keys that can be removed from the keyspace. */
+void activeExpireCycle(void) {
+ int j;
+
+ for (j = 0; j < server.dbnum; j++) {
+ int expired;
+ redisDb *db = server.db+j;
+
+ /* Continue to expire if at the end of the cycle more than 25%
+ * of the keys were expired. */
+ do {
+ long num = dictSize(db->expires);
+ time_t now = time(NULL);
+
+ expired = 0;
+ if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
+ num = REDIS_EXPIRELOOKUPS_PER_CRON;
+ while (num--) {
+ dictEntry *de;
+ time_t t;
+
+ if ((de = dictGetRandomKey(db->expires)) == NULL) break;
+ t = (time_t) dictGetEntryVal(de);
+ if (now > t) {
+ sds key = dictGetEntryKey(de);
+ robj *keyobj = createStringObject(key,sdslen(key));
+
+ propagateExpire(db,keyobj);
+ dbDelete(db,keyobj);
+ decrRefCount(keyobj);
+ expired++;
+ server.stat_expiredkeys++;
+ }
+ }
+ } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
+ }
+}
+
+void updateLRUClock(void) {
+ server.lruclock = (time(NULL)/REDIS_LRU_CLOCK_RESOLUTION) &
+ REDIS_LRU_CLOCK_MAX;
+}
+
int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
int j, loops = server.cronloops++;
REDIS_NOTUSED(eventLoop);
* in objects at every object access, and accuracy is not needed.
* To access a global var is faster than calling time(NULL) */
server.unixtime = time(NULL);
- /* We have just 21 bits per object for LRU information.
- * So we use an (eventually wrapping) LRU clock with minutes resolution.
+ /* We have just 22 bits per object for LRU information.
+ * So we use an (eventually wrapping) LRU clock with 10 seconds resolution.
+ * 2^22 bits with 10 seconds resoluton is more or less 1.5 years.
*
- * When we need to select what object to swap, we compute the minimum
- * time distance between the current lruclock and the object last access
- * lruclock info. Even if clocks will wrap on overflow, there is
- * the interesting property that we are sure that at least
- * ABS(A-B) minutes passed between current time and timestamp B.
+ * Note that even if this will wrap after 1.5 years it's not a problem,
+ * everything will still work but just some object will appear younger
+ * to Redis. But for this to happen a given object should never be touched
+ * for 1.5 years.
*
- * This is not precise but we don't need at all precision, but just
- * something statistically reasonable.
+ * Note that you can change the resolution altering the
+ * REDIS_LRU_CLOCK_RESOLUTION define.
*/
- server.lruclock = (time(NULL)/60)&((1<<21)-1);
+ updateLRUClock();
/* We received a SIGTERM, shutting down here in a safe way, as it is
* not ok doing so inside the signal handler. */
}
}
- /* Try to expire a few timed out keys. The algorithm used is adaptive and
- * will use few CPU cycles if there are few expiring keys, otherwise
- * it will get more aggressive to avoid that too much memory is used by
- * keys that can be removed from the keyspace. */
- for (j = 0; j < server.dbnum; j++) {
- int expired;
- redisDb *db = server.db+j;
-
- /* Continue to expire if at the end of the cycle more than 25%
- * of the keys were expired. */
- do {
- long num = dictSize(db->expires);
- time_t now = time(NULL);
-
- expired = 0;
- if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
- num = REDIS_EXPIRELOOKUPS_PER_CRON;
- while (num--) {
- dictEntry *de;
- time_t t;
-
- if ((de = dictGetRandomKey(db->expires)) == NULL) break;
- t = (time_t) dictGetEntryVal(de);
- if (now > t) {
- sds key = dictGetEntryKey(de);
- robj *keyobj = createStringObject(key,sdslen(key));
-
- dbDelete(db,keyobj);
- decrRefCount(keyobj);
- expired++;
- server.stat_expiredkeys++;
- }
- }
- } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
- }
+ /* Expire a few keys per cycle, only if this is a master.
+ * On slaves we wait for DEL operations synthesized by the master
+ * in order to guarantee a strict consistency. */
+ if (server.masterhost == NULL) activeExpireCycle();
/* Swap a few keys on disk if we are over the memory limit and VM
* is enbled. Try to free objects from the free list first. */
server.maxclients = 0;
server.blpop_blocked_clients = 0;
server.maxmemory = 0;
+ server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU;
+ server.maxmemory_samples = 3;
server.vm_enabled = 0;
server.vm_swap_file = zstrdup("/tmp/redis-%p.vm");
server.vm_page_size = 256; /* 256 bytes per page */
server.hash_max_zipmap_value = REDIS_HASH_MAX_ZIPMAP_VALUE;
server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES;
server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE;
+ server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES;
server.shutdown_asap = 0;
+ updateLRUClock();
resetServerSaveParams();
appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
server.stat_numconnections = 0;
server.stat_expiredkeys = 0;
server.stat_starttime = time(NULL);
+ server.stat_keyspace_misses = 0;
+ server.stat_keyspace_hits = 0;
server.unixtime = time(NULL);
aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
if (aeCreateFileEvent(server.el, server.fd, AE_READABLE,
void sortCommandTable() {
/* Copy and sort the read-only version of the command table */
- commandTable = (struct redisCommand*)malloc(sizeof(readonlyCommandTable));
+ commandTable = (struct redisCommand*)zmalloc(sizeof(readonlyCommandTable));
memcpy(commandTable,readonlyCommandTable,sizeof(readonlyCommandTable));
qsort(commandTable,
sizeof(readonlyCommandTable)/sizeof(struct redisCommand),
int processCommand(redisClient *c) {
struct redisCommand *cmd;
- /* Free some memory if needed (maxmemory setting) */
- if (server.maxmemory) freeMemoryIfNeeded();
-
/* Handle the multi bulk command type. This is an alternative protocol
* supported by Redis in order to receive commands that are composed of
* multiple binary-safe "bulk" arguments. The latency of processing is
} else if (c->multibulk) {
if (c->bulklen == -1) {
if (((char*)c->argv[0]->ptr)[0] != '$') {
- addReplySds(c,sdsnew("-ERR multi bulk protocol error\r\n"));
+ addReplyError(c,"multi bulk protocol error");
resetClient(c);
return 1;
} else {
- int bulklen = atoi(((char*)c->argv[0]->ptr)+1);
+ char *eptr;
+ long bulklen = strtol(((char*)c->argv[0]->ptr)+1,&eptr,10);
+ int perr = eptr[0] != '\0';
+
decrRefCount(c->argv[0]);
- if (bulklen < 0 || bulklen > 1024*1024*1024) {
+ if (perr || bulklen == LONG_MIN || bulklen == LONG_MAX ||
+ bulklen < 0 || bulklen > 1024*1024*1024)
+ {
c->argc--;
- addReplySds(c,sdsnew("-ERR invalid bulk write count\r\n"));
+ addReplyError(c,"invalid bulk write count");
resetClient(c);
return 1;
}
* such wrong arity, bad command name and so forth. */
cmd = lookupCommand(c->argv[0]->ptr);
if (!cmd) {
- addReplySds(c,
- sdscatprintf(sdsempty(), "-ERR unknown command '%s'\r\n",
- (char*)c->argv[0]->ptr));
+ addReplyErrorFormat(c,"unknown command '%s'",
+ (char*)c->argv[0]->ptr);
resetClient(c);
return 1;
} else if ((cmd->arity > 0 && cmd->arity != c->argc) ||
(c->argc < -cmd->arity)) {
- addReplySds(c,
- sdscatprintf(sdsempty(),
- "-ERR wrong number of arguments for '%s' command\r\n",
- cmd->name));
+ addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
+ cmd->name);
resetClient(c);
return 1;
} else if (cmd->flags & REDIS_CMD_BULK && c->bulklen == -1) {
/* This is a bulk command, we have to read the last argument yet. */
- int bulklen = atoi(c->argv[c->argc-1]->ptr);
+ char *eptr;
+ long bulklen = strtol(c->argv[c->argc-1]->ptr,&eptr,10);
+ int perr = eptr[0] != '\0';
decrRefCount(c->argv[c->argc-1]);
- if (bulklen < 0 || bulklen > 1024*1024*1024) {
+ if (perr || bulklen == LONG_MAX || bulklen == LONG_MIN ||
+ bulklen < 0 || bulklen > 1024*1024*1024)
+ {
c->argc--;
- addReplySds(c,sdsnew("-ERR invalid bulk write count\r\n"));
+ addReplyError(c,"invalid bulk write count");
resetClient(c);
return 1;
}
/* Check if the user is authenticated */
if (server.requirepass && !c->authenticated && cmd->proc != authCommand) {
- addReplySds(c,sdsnew("-ERR operation not permitted\r\n"));
+ addReplyError(c,"operation not permitted");
resetClient(c);
return 1;
}
- /* Handle the maxmemory directive */
+ /* Handle the maxmemory directive.
+ *
+ * First we try to free some memory if possible (if there are volatile
+ * keys in the dataset). If there are not the only thing we can do
+ * is returning an error. */
+ if (server.maxmemory) freeMemoryIfNeeded();
if (server.maxmemory && (cmd->flags & REDIS_CMD_DENYOOM) &&
zmalloc_used_memory() > server.maxmemory)
{
- addReplySds(c,sdsnew("-ERR command not allowed when used memory > 'maxmemory'\r\n"));
+ addReplyError(c,"command not allowed when used memory > 'maxmemory'");
resetClient(c);
return 1;
}
&&
cmd->proc != subscribeCommand && cmd->proc != unsubscribeCommand &&
cmd->proc != psubscribeCommand && cmd->proc != punsubscribeCommand) {
- addReplySds(c,sdsnew("-ERR only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context\r\n"));
+ addReplyError(c,"only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context");
resetClient(c);
return 1;
}
/* Append only file: fsync() the AOF and exit */
aof_fsync(server.appendfd);
if (server.vm_enabled) unlink(server.vm_swap_file);
- } else {
+ } else if (server.saveparamslen > 0) {
/* Snapshotting. Perform a SYNC SAVE and exit */
- if (rdbSave(server.dbfilename) == REDIS_OK) {
- if (server.daemonize)
- unlink(server.pidfile);
- redisLog(REDIS_WARNING,"%zu bytes used at exit",zmalloc_used_memory());
- } else {
+ if (rdbSave(server.dbfilename) != REDIS_OK) {
/* Ooops.. error saving! The best we can do is to continue
* operating. Note that if there was a background saving process,
* in the next cron() Redis will be notified that the background
redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit");
return REDIS_ERR;
}
+ } else {
+ redisLog(REDIS_WARNING,"Not saving DB.");
}
+ if (server.daemonize) unlink(server.pidfile);
redisLog(REDIS_WARNING,"Server exit now, bye bye...");
return REDIS_OK;
}
addReply(c,shared.ok);
} else {
c->authenticated = 0;
- addReplySds(c,sdscatprintf(sdsempty(),"-ERR invalid password\r\n"));
+ addReplyError(c,"invalid password");
}
}
time_t uptime = time(NULL)-server.stat_starttime;
int j;
char hmem[64];
+ struct rusage self_ru, c_ru;
+
+ getrusage(RUSAGE_SELF, &self_ru);
+ getrusage(RUSAGE_CHILDREN, &c_ru);
bytesToHuman(hmem,zmalloc_used_memory());
info = sdscatprintf(sdsempty(),
"process_id:%ld\r\n"
"uptime_in_seconds:%ld\r\n"
"uptime_in_days:%ld\r\n"
+ "lru_clock:%ld\r\n"
+ "used_cpu_sys:%.2f\r\n"
+ "used_cpu_user:%.2f\r\n"
+ "used_cpu_sys_childrens:%.2f\r\n"
+ "used_cpu_user_childrens:%.2f\r\n"
"connected_clients:%d\r\n"
"connected_slaves:%d\r\n"
"blocked_clients:%d\r\n"
"used_memory:%zu\r\n"
"used_memory_human:%s\r\n"
+ "mem_fragmentation_ratio:%.2f\r\n"
+ "use_tcmalloc:%d\r\n"
"changes_since_last_save:%lld\r\n"
"bgsave_in_progress:%d\r\n"
"last_save_time:%ld\r\n"
"total_connections_received:%lld\r\n"
"total_commands_processed:%lld\r\n"
"expired_keys:%lld\r\n"
+ "keyspace_hits:%lld\r\n"
+ "keyspace_misses:%lld\r\n"
"hash_max_zipmap_entries:%zu\r\n"
"hash_max_zipmap_value:%zu\r\n"
"pubsub_channels:%ld\r\n"
(long) getpid(),
uptime,
uptime/(3600*24),
+ (unsigned long) server.lruclock,
+ (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000,
+ (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000,
+ (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000,
+ (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000,
listLength(server.clients)-listLength(server.slaves),
listLength(server.slaves),
server.blpop_blocked_clients,
zmalloc_used_memory(),
hmem,
+ zmalloc_get_fragmentation_ratio(),
+#ifdef USE_TCMALLOC
+ 1,
+#else
+ 0,
+#endif
server.dirty,
server.bgsavechildpid != -1,
server.lastsave,
server.stat_numconnections,
server.stat_numcommands,
server.stat_expiredkeys,
+ server.stat_keyspace_hits,
+ server.stat_keyspace_misses,
server.hash_max_zipmap_entries,
server.hash_max_zipmap_value,
dictSize(server.pubsub_channels),
* memory usage.
*/
void freeMemoryIfNeeded(void) {
+ /* Remove keys accordingly to the active policy as long as we are
+ * over the memory limit. */
while (server.maxmemory && zmalloc_used_memory() > server.maxmemory) {
int j, k, freed = 0;
+ /* Basic strategy -- remove objects from the free list. */
if (tryFreeOneObjectFromFreelist() == REDIS_OK) continue;
+
+ for (j = 0; j < server.dbnum; j++) {
+ long bestval;
+ sds bestkey = NULL;
+ struct dictEntry *de;
+ redisDb *db = server.db+j;
+ dict *dict;
+
+ if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
+ server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM)
+ {
+ dict = server.db[j].dict;
+ } else {
+ dict = server.db[j].expires;
+ }
+ if (dictSize(dict) == 0) continue;
+
+ /* volatile-random and allkeys-random policy */
+ if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_RANDOM ||
+ server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_RANDOM)
+ {
+ de = dictGetRandomKey(dict);
+ bestkey = dictGetEntryKey(de);
+ }
+
+ /* volatile-lru and allkeys-lru policy */
+ else if (server.maxmemory_policy == REDIS_MAXMEMORY_ALLKEYS_LRU ||
+ server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU)
+ {
+ for (k = 0; k < server.maxmemory_samples; k++) {
+ sds thiskey;
+ long thisval;
+ robj *o;
+
+ de = dictGetRandomKey(dict);
+ thiskey = dictGetEntryKey(de);
+ o = dictGetEntryVal(de);
+ thisval = estimateObjectIdleTime(o);
+
+ /* Higher idle time is better candidate for deletion */
+ if (bestkey == NULL || thisval > bestval) {
+ bestkey = thiskey;
+ bestval = thisval;
+ }
+ }
+ }
+
+ /* volatile-ttl */
+ else if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_TTL) {
+ for (k = 0; k < server.maxmemory_samples; k++) {
+ sds thiskey;
+ long thisval;
+
+ de = dictGetRandomKey(dict);
+ thiskey = dictGetEntryKey(de);
+ thisval = (long) dictGetEntryVal(de);
+
+ /* Expire sooner (minor expire unix timestamp) is better
+ * candidate for deletion */
+ if (bestkey == NULL || thisval < bestval) {
+ bestkey = thiskey;
+ bestval = thisval;
+ }
+ }
+ }
+
+ /* Finally remove the selected key. */
+ if (bestkey) {
+ robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
+ dbDelete(db,keyobj);
+ server.stat_expiredkeys++;
+ decrRefCount(keyobj);
+ freed++;
+ }
+ }
+ if (!freed) return; /* nothing to free... */
+ }
+
+ while(0) {
+ int j, k, freed = 0;
for (j = 0; j < server.dbnum; j++) {
int minttl = -1;
- robj *minkey = NULL;
+ sds minkey = NULL;
+ robj *keyobj = NULL;
struct dictEntry *de;
if (dictSize(server.db[j].expires)) {
minttl = t;
}
}
- dbDelete(server.db+j,minkey);
+ keyobj = createStringObject(minkey,sdslen(minkey));
+ dbDelete(server.db+j,keyobj);
+ server.stat_expiredkeys++;
+ decrRefCount(keyobj);
}
}
if (!freed) return; /* nothing to free... */
}
#endif /* __linux__ */
+void createPidFile(void) {
+ /* Try to write the pid file in a best-effort way. */
+ FILE *fp = fopen(server.pidfile,"w");
+ if (fp) {
+ fprintf(fp,"%d\n",getpid());
+ fclose(fp);
+ }
+}
+
void daemonize(void) {
int fd;
- FILE *fp;
if (fork() != 0) exit(0); /* parent exits */
setsid(); /* create a new session */
dup2(fd, STDERR_FILENO);
if (fd > STDERR_FILENO) close(fd);
}
- /* Try to write the pid file */
- fp = fopen(server.pidfile,"w");
- if (fp) {
- fprintf(fp,"%d\n",getpid());
- fclose(fp);
- }
}
void version() {
}
if (server.daemonize) daemonize();
initServer();
+ if (server.daemonize) createPidFile();
redisLog(REDIS_NOTICE,"Server started, Redis version " REDIS_VERSION);
#ifdef __linux__
linuxOvercommitMemoryWarning();
redisLog(REDIS_WARNING,"%s", messages[i]);
/* free(messages); Don't call free() with possibly corrupted memory. */
+ if (server.daemonize) unlink(server.pidfile);
_exit(0);
}
projects / redis.git / blobdiff