Make replication faster (biggest gain for small number of slaves)

[redis.git] / src / object.c

diff --git a/src/object.c b/src/object.c
index 429ac0ecc6cd7eea9cabd7b987701dd9153278a0..22f538371a72bca93c4bd1095febf12a98766c66 100644 (file)
--- a/src/object.c
+++ b/src/object.c
@@ -3,30 +3,24 @@
 #include <math.h>
 
 robj *createObject(int type, void *ptr) {
 #include <math.h>
 
 robj *createObject(int type, void *ptr) {

-    robj *o;
-
-    if (server.vm_enabled) pthread_mutex_lock(&server.obj_freelist_mutex);
-    if (listLength(server.objfreelist)) {
-        listNode *head = listFirst(server.objfreelist);
-        o = listNodeValue(head);
-        listDelNode(server.objfreelist,head);
-        if (server.vm_enabled) pthread_mutex_unlock(&server.obj_freelist_mutex);
-    } else {
-        if (server.vm_enabled) pthread_mutex_unlock(&server.obj_freelist_mutex);
-        o = zmalloc(sizeof(*o));
-    }
+    robj *o = zmalloc(sizeof(*o));

     o->type = type;
     o->encoding = REDIS_ENCODING_RAW;
     o->ptr = ptr;
     o->refcount = 1;
     o->type = type;
     o->encoding = REDIS_ENCODING_RAW;
     o->ptr = ptr;
     o->refcount = 1;

-    if (server.vm_enabled) {
-        /* Note that this code may run in the context of an I/O thread
-         * and accessing server.lruclock in theory is an error
-         * (no locks). But in practice this is safe, and even if we read
-         * garbage Redis will not fail. */
-        o->lru = server.lruclock;
-        o->storage = REDIS_VM_MEMORY;
-    }
+
+    /* Set the LRU to the current lruclock (minutes resolution).
+     * We do this regardless of the fact VM is active as LRU is also
+     * used for the maxmemory directive when Redis is used as cache.
+     *
+     * Note that this code may run in the context of an I/O thread
+     * and accessing server.lruclock in theory is an error
+     * (no locks). But in practice this is safe, and even if we read
+     * garbage Redis will not fail. */
+    o->lru = server.lruclock;
+    /* The following is only needed if VM is active, but since the conditional
+     * is probably more costly than initializing the field it's better to
+     * have every field properly initialized anyway. */

     return o;
 }
 
     return o;
 }
 


@@ -37,6 +31,7 @@ robj *createStringObject(char *ptr, size_t len) {
 robj *createStringObjectFromLongLong(long long value) {
     robj *o;
     if (value >= 0 && value < REDIS_SHARED_INTEGERS &&
 robj *createStringObjectFromLongLong(long long value) {
     robj *o;
     if (value >= 0 && value < REDIS_SHARED_INTEGERS &&

+        !server.ds_enabled &&

         pthread_equal(pthread_self(),server.mainthread)) {
         incrRefCount(shared.integers[value]);
         o = shared.integers[value];
         pthread_equal(pthread_self(),server.mainthread)) {
         incrRefCount(shared.integers[value]);
         o = shared.integers[value];


@@ -74,7 +69,16 @@ robj *createZiplistObject(void) {
 
 robj *createSetObject(void) {
     dict *d = dictCreate(&setDictType,NULL);
 
 robj *createSetObject(void) {
     dict *d = dictCreate(&setDictType,NULL);

-    return createObject(REDIS_SET,d);
+    robj *o = createObject(REDIS_SET,d);
+    o->encoding = REDIS_ENCODING_HT;
+    return o;
+}
+
+robj *createIntsetObject(void) {
+    intset *is = intsetNew();
+    robj *o = createObject(REDIS_SET,is);
+    o->encoding = REDIS_ENCODING_INTSET;
+    return o;

 }
 
 robj *createHashObject(void) {
 }
 
 robj *createHashObject(void) {


@@ -89,10 +93,20 @@ robj *createHashObject(void) {
 
 robj *createZsetObject(void) {
     zset *zs = zmalloc(sizeof(*zs));
 
 robj *createZsetObject(void) {
     zset *zs = zmalloc(sizeof(*zs));

+    robj *o;

 
     zs->dict = dictCreate(&zsetDictType,NULL);
     zs->zsl = zslCreate();
 
     zs->dict = dictCreate(&zsetDictType,NULL);
     zs->zsl = zslCreate();

-    return createObject(REDIS_ZSET,zs);
+    o = createObject(REDIS_ZSET,zs);
+    o->encoding = REDIS_ENCODING_SKIPLIST;
+    return o;
+}
+
+robj *createZsetZiplistObject(void) {
+    unsigned char *zl = ziplistNew();
+    robj *o = createObject(REDIS_ZSET,zl);
+    o->encoding = REDIS_ENCODING_ZIPLIST;
+    return o;

 }
 
 void freeStringObject(robj *o) {
 }
 
 void freeStringObject(robj *o) {


@@ -115,15 +129,33 @@ void freeListObject(robj *o) {
 }
 
 void freeSetObject(robj *o) {
 }
 
 void freeSetObject(robj *o) {

-    dictRelease((dict*) o->ptr);
+    switch (o->encoding) {
+    case REDIS_ENCODING_HT:
+        dictRelease((dict*) o->ptr);
+        break;
+    case REDIS_ENCODING_INTSET:
+        zfree(o->ptr);
+        break;
+    default:
+        redisPanic("Unknown set encoding type");
+    }

 }
 
 void freeZsetObject(robj *o) {
 }
 
 void freeZsetObject(robj *o) {

-    zset *zs = o->ptr;
-
-    dictRelease(zs->dict);
-    zslFree(zs->zsl);
-    zfree(zs);
+    zset *zs;
+    switch (o->encoding) {
+    case REDIS_ENCODING_SKIPLIST:
+        zs = o->ptr;
+        dictRelease(zs->dict);
+        zslFree(zs->zsl);
+        zfree(zs);
+        break;
+    case REDIS_ENCODING_ZIPLIST:
+        zfree(o->ptr);
+        break;
+    default:
+        redisPanic("Unknown sorted set encoding");
+    }

 }
 
 void freeHashObject(robj *o) {
 }
 
 void freeHashObject(robj *o) {


@@ -147,30 +179,7 @@ void incrRefCount(robj *o) {
 void decrRefCount(void *obj) {
     robj *o = obj;
 
 void decrRefCount(void *obj) {
     robj *o = obj;
 

-    /* Object is a swapped out value, or in the process of being loaded. */
-    if (server.vm_enabled &&
-        (o->storage == REDIS_VM_SWAPPED || o->storage == REDIS_VM_LOADING))
-    {
-        vmpointer *vp = obj;
-        if (o->storage == REDIS_VM_LOADING) vmCancelThreadedIOJob(o);
-        vmMarkPagesFree(vp->page,vp->usedpages);
-        server.vm_stats_swapped_objects--;
-        zfree(vp);
-        return;
-    }
-

     if (o->refcount <= 0) redisPanic("decrRefCount against refcount <= 0");
     if (o->refcount <= 0) redisPanic("decrRefCount against refcount <= 0");

-    /* Object is in memory, or in the process of being swapped out.
-     *
-     * If the object is being swapped out, abort the operation on
-     * decrRefCount even if the refcount does not drop to 0: the object
-     * is referenced at least two times, as value of the key AND as
-     * job->val in the iojob. So if we don't invalidate the iojob, when it is
-     * done but the relevant key was removed in the meantime, the
-     * complete jobs handler will not find the key about the job and the
-     * assert will fail. */
-    if (server.vm_enabled && o->storage == REDIS_VM_SWAPPING)
-        vmCancelThreadedIOJob(o);

     if (--(o->refcount) == 0) {
         switch(o->type) {
         case REDIS_STRING: freeStringObject(o); break;
     if (--(o->refcount) == 0) {
         switch(o->type) {
         case REDIS_STRING: freeStringObject(o); break;


@@ -181,11 +190,7 @@ void decrRefCount(void *obj) {
         default: redisPanic("Unknown object type"); break;
         }
         o->ptr = NULL; /* defensive programming. We'll see NULL in traces. */
         default: redisPanic("Unknown object type"); break;
         }
         o->ptr = NULL; /* defensive programming. We'll see NULL in traces. */

-        if (server.vm_enabled) pthread_mutex_lock(&server.obj_freelist_mutex);
-        if (listLength(server.objfreelist) > REDIS_OBJFREELIST_MAX ||
-            !listAddNodeHead(server.objfreelist,o))
-            zfree(o);
-        if (server.vm_enabled) pthread_mutex_unlock(&server.obj_freelist_mutex);
+        zfree(o);

     }
 }
 
     }
 }
 


@@ -197,6 +202,16 @@ int checkType(redisClient *c, robj *o, int type) {
     return 0;
 }
 
     return 0;
 }
 

+int isObjectRepresentableAsLongLong(robj *o, long long *llval) {
+    redisAssert(o->type == REDIS_STRING);
+    if (o->encoding == REDIS_ENCODING_INT) {
+        if (llval) *llval = (long) o->ptr;
+        return REDIS_OK;
+    } else {
+        return string2ll(o->ptr,sdslen(o->ptr),llval) ? REDIS_OK : REDIS_ERR;
+    }
+}
+

 /* Try to encode a string object in order to save space */
 robj *tryObjectEncoding(robj *o) {
     long value;
 /* Try to encode a string object in order to save space */
 robj *tryObjectEncoding(robj *o) {
     long value;


@@ -214,17 +229,21 @@ robj *tryObjectEncoding(robj *o) {
     redisAssert(o->type == REDIS_STRING);
 
     /* Check if we can represent this string as a long integer */
     redisAssert(o->type == REDIS_STRING);
 
     /* Check if we can represent this string as a long integer */

-    if (isStringRepresentableAsLong(s,&value) == REDIS_ERR) return o;
+    if (!string2l(s,sdslen(s),&value)) return o;

 
     /* Ok, this object can be encoded...
      *
      * Can I use a shared object? Only if the object is inside a given
 
     /* Ok, this object can be encoded...
      *
      * Can I use a shared object? Only if the object is inside a given

-     * range and if this is the main thread, since when VM is enabled we
-     * have the constraint that I/O thread should only handle non-shared
-     * objects, in order to avoid race conditions (we don't have per-object
-     * locking). */
-    if (value >= 0 && value < REDIS_SHARED_INTEGERS &&
-        pthread_equal(pthread_self(),server.mainthread)) {
+     * range and if the back end in use is in-memory. For disk store every
+     * object in memory used as value should be independent.
+     *
+     * Note that we also avoid using shared integers when maxmemory is used
+     * because every object needs to have a private LRU field for the LRU
+     * algorithm to work well. */
+    if (!server.ds_enabled &&
+        server.maxmemory == 0 && value >= 0 && value < REDIS_SHARED_INTEGERS &&
+        pthread_equal(pthread_self(),server.mainthread))
+    {

         decrRefCount(o);
         incrRefCount(shared.integers[value]);
         return shared.integers[value];
         decrRefCount(o);
         incrRefCount(shared.integers[value]);
         return shared.integers[value];


@@ -336,9 +355,9 @@ int getDoubleFromObjectOrReply(redisClient *c, robj *o, double *target, const ch
     double value;
     if (getDoubleFromObject(o, &value) != REDIS_OK) {
         if (msg != NULL) {
     double value;
     if (getDoubleFromObject(o, &value) != REDIS_OK) {
         if (msg != NULL) {

-            addReplySds(c, sdscatprintf(sdsempty(), "-ERR %s\r\n", msg));
+            addReplyError(c,(char*)msg);

         } else {
         } else {

-            addReplySds(c, sdsnew("-ERR value is not a double\r\n"));
+            addReplyError(c,"value is not a double");

         }
         return REDIS_ERR;
     }
         }
         return REDIS_ERR;
     }


@@ -367,7 +386,7 @@ int getLongLongFromObject(robj *o, long long *target) {
         }
     }
 
         }
     }
 

-    *target = value;
+    if (target) *target = value;

     return REDIS_OK;
 }
 
     return REDIS_OK;
 }
 


@@ -375,9 +394,9 @@ int getLongLongFromObjectOrReply(redisClient *c, robj *o, long long *target, con
     long long value;
     if (getLongLongFromObject(o, &value) != REDIS_OK) {
         if (msg != NULL) {
     long long value;
     if (getLongLongFromObject(o, &value) != REDIS_OK) {
         if (msg != NULL) {

-            addReplySds(c, sdscatprintf(sdsempty(), "-ERR %s\r\n", msg));
+            addReplyError(c,(char*)msg);

         } else {
         } else {

-            addReplySds(c, sdsnew("-ERR value is not an integer or out of range\r\n"));
+            addReplyError(c,"value is not an integer or out of range");

         }
         return REDIS_ERR;
     }
         }
         return REDIS_ERR;
     }


@@ -392,9 +411,9 @@ int getLongFromObjectOrReply(redisClient *c, robj *o, long *target, const char *
     if (getLongLongFromObjectOrReply(c, o, &value, msg) != REDIS_OK) return REDIS_ERR;
     if (value < LONG_MIN || value > LONG_MAX) {
         if (msg != NULL) {
     if (getLongLongFromObjectOrReply(c, o, &value, msg) != REDIS_OK) return REDIS_ERR;
     if (value < LONG_MIN || value > LONG_MAX) {
         if (msg != NULL) {

-            addReplySds(c, sdscatprintf(sdsempty(), "-ERR %s\r\n", msg));
+            addReplyError(c,(char*)msg);

         } else {
         } else {

-            addReplySds(c, sdsnew("-ERR value is out of range\r\n"));
+            addReplyError(c,"value is out of range");

         }
         return REDIS_ERR;
     }
         }
         return REDIS_ERR;
     }


@@ -411,6 +430,58 @@ char *strEncoding(int encoding) {
     case REDIS_ENCODING_ZIPMAP: return "zipmap";
     case REDIS_ENCODING_LINKEDLIST: return "linkedlist";
     case REDIS_ENCODING_ZIPLIST: return "ziplist";
     case REDIS_ENCODING_ZIPMAP: return "zipmap";
     case REDIS_ENCODING_LINKEDLIST: return "linkedlist";
     case REDIS_ENCODING_ZIPLIST: return "ziplist";

+    case REDIS_ENCODING_INTSET: return "intset";
+    case REDIS_ENCODING_SKIPLIST: return "skiplist";

     default: return "unknown";
     }
 }
     default: return "unknown";
     }
 }

+
+/* Given an object returns the min number of seconds the object was never
+ * requested, using an approximated LRU algorithm. */
+unsigned long estimateObjectIdleTime(robj *o) {
+    if (server.lruclock >= o->lru) {
+        return (server.lruclock - o->lru) * REDIS_LRU_CLOCK_RESOLUTION;
+    } else {
+        return ((REDIS_LRU_CLOCK_MAX - o->lru) + server.lruclock) *
+                    REDIS_LRU_CLOCK_RESOLUTION;
+    }
+}
+
+/* This is an helper function for the DEBUG command. We need to lookup keys
+ * without any modification of LRU or other parameters. */
+robj *objectCommandLookup(redisClient *c, robj *key) {
+    dictEntry *de;
+
+    if ((de = dictFind(c->db->dict,key->ptr)) == NULL) return NULL;
+    return (robj*) dictGetEntryVal(de);
+}
+
+robj *objectCommandLookupOrReply(redisClient *c, robj *key, robj *reply) {
+    robj *o = objectCommandLookup(c,key);
+
+    if (!o) addReply(c, reply);
+    return o;
+}
+
+/* Object command allows to inspect the internals of an Redis Object.
+ * Usage: OBJECT <verb> ... arguments ... */
+void objectCommand(redisClient *c) {
+    robj *o;
+
+    if (!strcasecmp(c->argv[1]->ptr,"refcount") && c->argc == 3) {
+        if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.nullbulk))
+                == NULL) return;
+        addReplyLongLong(c,o->refcount);
+    } else if (!strcasecmp(c->argv[1]->ptr,"encoding") && c->argc == 3) {
+        if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.nullbulk))
+                == NULL) return;
+        addReplyBulkCString(c,strEncoding(o->encoding));
+    } else if (!strcasecmp(c->argv[1]->ptr,"idletime") && c->argc == 3) {
+        if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.nullbulk))
+                == NULL) return;
+        addReplyLongLong(c,estimateObjectIdleTime(o));
+    } else {
+        addReplyError(c,"Syntax error. Try OBJECT (refcount|encoding|idletime)");
+    }
+}
+