direct saving of specially encoded types implemented for lists and sets too

[redis.git] / src / rdb.c

diff --git a/src/rdb.c b/src/rdb.c
index 5e69a32443a31f673b9a825dc80492a739fbc15b..fcf97d335a580d69e20a8c983e4c516392c6a9d2 100644 (file)
--- a/src/rdb.c
+++ b/src/rdb.c
@@ -139,7 +139,7 @@ writeerr:
 }
 
 /* Save a string objet as [len][data] on disk. If the object is a string
 }
 
 /* Save a string objet as [len][data] on disk. If the object is a string

- * representation of an integer value we try to safe it in a special form */
+ * representation of an integer value we try to save it in a special form */

 int rdbSaveRawString(FILE *fp, unsigned char *s, size_t len) {
     int enclen;
     int n, nwritten = 0;
 int rdbSaveRawString(FILE *fp, unsigned char *s, size_t len) {
     int enclen;
     int n, nwritten = 0;


@@ -256,27 +256,10 @@ int rdbSaveObject(FILE *fp, robj *o) {
     } else if (o->type == REDIS_LIST) {
         /* Save a list value */
         if (o->encoding == REDIS_ENCODING_ZIPLIST) {
     } else if (o->type == REDIS_LIST) {
         /* Save a list value */
         if (o->encoding == REDIS_ENCODING_ZIPLIST) {

-            unsigned char *p;
-            unsigned char *vstr;
-            unsigned int vlen;
-            long long vlong;
+            size_t l = ziplistBlobLen((unsigned char*)o->ptr);

 
 

-            if ((n = rdbSaveLen(fp,ziplistLen(o->ptr))) == -1) return -1;
+            if ((n = rdbSaveRawString(fp,o->ptr,l)) == -1) return -1;

             nwritten += n;
             nwritten += n;

-
-            p = ziplistIndex(o->ptr,0);
-            while(ziplistGet(p,&vstr,&vlen,&vlong)) {
-                if (vstr) {
-                    if ((n = rdbSaveRawString(fp,vstr,vlen)) == -1)
-                        return -1;
-                    nwritten += n;
-                } else {
-                    if ((n = rdbSaveLongLongAsStringObject(fp,vlong)) == -1)
-                        return -1;
-                    nwritten += n;
-                }
-                p = ziplistNext(o->ptr,p);
-            }

         } else if (o->encoding == REDIS_ENCODING_LINKEDLIST) {
             list *list = o->ptr;
             listIter li;
         } else if (o->encoding == REDIS_ENCODING_LINKEDLIST) {
             list *list = o->ptr;
             listIter li;


@@ -311,17 +294,10 @@ int rdbSaveObject(FILE *fp, robj *o) {
             }
             dictReleaseIterator(di);
         } else if (o->encoding == REDIS_ENCODING_INTSET) {
             }
             dictReleaseIterator(di);
         } else if (o->encoding == REDIS_ENCODING_INTSET) {

-            intset *is = o->ptr;
-            int64_t llval;
-            int i = 0;
+            size_t l = intsetBlobLen((intset*)o->ptr);

 
 

-            if ((n = rdbSaveLen(fp,intsetLen(is))) == -1) return -1;
+            if ((n = rdbSaveRawString(fp,o->ptr,l)) == -1) return -1;

             nwritten += n;
             nwritten += n;

-
-            while(intsetGet(is,i++,&llval)) {
-                if ((n = rdbSaveLongLongAsStringObject(fp,llval)) == -1) return -1;
-                nwritten += n;
-            }

         } else {
             redisPanic("Unknown set encoding");
         }
         } else {
             redisPanic("Unknown set encoding");
         }


@@ -347,20 +323,10 @@ int rdbSaveObject(FILE *fp, robj *o) {
     } else if (o->type == REDIS_HASH) {
         /* Save a hash value */
         if (o->encoding == REDIS_ENCODING_ZIPMAP) {
     } else if (o->type == REDIS_HASH) {
         /* Save a hash value */
         if (o->encoding == REDIS_ENCODING_ZIPMAP) {

-            unsigned char *p = zipmapRewind(o->ptr);
-            unsigned int count = zipmapLen(o->ptr);
-            unsigned char *key, *val;
-            unsigned int klen, vlen;
+            size_t l = zipmapBlobLen((unsigned char*)o->ptr);

 
 

-            if ((n = rdbSaveLen(fp,count)) == -1) return -1;
+            if ((n = rdbSaveRawString(fp,o->ptr,l)) == -1) return -1;

             nwritten += n;
             nwritten += n;

-
-            while((p = zipmapNext(p,&key,&klen,&val,&vlen)) != NULL) {
-                if ((n = rdbSaveRawString(fp,key,klen)) == -1) return -1;
-                nwritten += n;
-                if ((n = rdbSaveRawString(fp,val,vlen)) == -1) return -1;
-                nwritten += n;
-            }

         } else {
             dictIterator *di = dictGetIterator(o->ptr);
             dictEntry *de;
         } else {
             dictIterator *di = dictGetIterator(o->ptr);
             dictEntry *de;


@@ -395,10 +361,34 @@ off_t rdbSavedObjectLen(robj *o) {
     return len;
 }
 
     return len;
 }
 

-/* Return the number of pages required to save this object in the swap file */
-off_t rdbSavedObjectPages(robj *o) {
-    off_t bytes = rdbSavedObjectLen(o);
-    return (bytes+(server.vm_page_size-1))/server.vm_page_size;
+/* Save a key-value pair, with expire time, type, key, value.
+ * On error -1 is returned.
+ * On success if the key was actaully saved 1 is returned, otherwise 0
+ * is returned (the key was already expired). */
+int rdbSaveKeyValuePair(FILE *fp, robj *key, robj *val,
+                        time_t expiretime, time_t now)
+{
+    int vtype;
+
+    /* Save the expire time */
+    if (expiretime != -1) {
+        /* If this key is already expired skip it */
+        if (expiretime < now) return 0;
+        if (rdbSaveType(fp,REDIS_EXPIRETIME) == -1) return -1;
+        if (rdbSaveTime(fp,expiretime) == -1) return -1;
+    }
+    /* Fix the object type if needed, to support saving zipmaps, ziplists,
+     * and intsets, directly as blobs of bytes: they are already serialized. */
+    vtype = val->type;
+    if (vtype == REDIS_HASH && val->encoding == REDIS_ENCODING_ZIPMAP)
+        vtype = REDIS_HASH_ZIPMAP;
+    else if (vtype == REDIS_LIST && val->encoding == REDIS_ENCODING_ZIPLIST)
+        vtype = REDIS_LIST_ZIPLIST;
+    /* Save type, key, value */
+    if (rdbSaveType(fp,vtype) == -1) return -1;
+    if (rdbSaveStringObject(fp,key) == -1) return -1;
+    if (rdbSaveObject(fp,val) == -1) return -1;
+    return 1;

 }
 
 /* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success */
 }
 
 /* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success */


@@ -410,16 +400,16 @@ int rdbSave(char *filename) {
     int j;
     time_t now = time(NULL);
 
     int j;
     time_t now = time(NULL);
 

-    /* Wait for I/O therads to terminate, just in case this is a
-     * foreground-saving, to avoid seeking the swap file descriptor at the
-     * same time. */
-    if (server.vm_enabled)
-        waitEmptyIOJobsQueue();
+    if (server.ds_enabled) {
+        cacheForcePointInTime();
+        return dsRdbSave(filename);
+    }

 
     snprintf(tmpfile,256,"temp-%d.rdb", (int) getpid());
     fp = fopen(tmpfile,"w");
     if (!fp) {
 
     snprintf(tmpfile,256,"temp-%d.rdb", (int) getpid());
     fp = fopen(tmpfile,"w");
     if (!fp) {

-        redisLog(REDIS_WARNING, "Failed saving the DB: %s", strerror(errno));
+        redisLog(REDIS_WARNING, "Failed opening .rdb for saving: %s",
+            strerror(errno));

         return REDIS_ERR;
     }
     if (fwrite("REDIS0001",9,1,fp) == 0) goto werr;
         return REDIS_ERR;
     }
     if (fwrite("REDIS0001",9,1,fp) == 0) goto werr;


@@ -441,38 +431,11 @@ int rdbSave(char *filename) {
         while((de = dictNext(di)) != NULL) {
             sds keystr = dictGetEntryKey(de);
             robj key, *o = dictGetEntryVal(de);
         while((de = dictNext(di)) != NULL) {
             sds keystr = dictGetEntryKey(de);
             robj key, *o = dictGetEntryVal(de);

-            time_t expiretime;
+            time_t expire;

             
             initStaticStringObject(key,keystr);
             
             initStaticStringObject(key,keystr);

-            expiretime = getExpire(db,&key);
-
-            /* Save the expire time */
-            if (expiretime != -1) {
-                /* If this key is already expired skip it */
-                if (expiretime < now) continue;
-                if (rdbSaveType(fp,REDIS_EXPIRETIME) == -1) goto werr;
-                if (rdbSaveTime(fp,expiretime) == -1) goto werr;
-            }
-            /* Save the key and associated value. This requires special
-             * handling if the value is swapped out. */
-            if (!server.vm_enabled || o->storage == REDIS_VM_MEMORY ||
-                                      o->storage == REDIS_VM_SWAPPING) {
-                /* Save type, key, value */
-                if (rdbSaveType(fp,o->type) == -1) goto werr;
-                if (rdbSaveStringObject(fp,&key) == -1) goto werr;
-                if (rdbSaveObject(fp,o) == -1) goto werr;
-            } else {
-                /* REDIS_VM_SWAPPED or REDIS_VM_LOADING */
-                robj *po;
-                /* Get a preview of the object in memory */
-                po = vmPreviewObject(o);
-                /* Save type, key, value */
-                if (rdbSaveType(fp,po->type) == -1) goto werr;
-                if (rdbSaveStringObject(fp,&key) == -1) goto werr;
-                if (rdbSaveObject(fp,po) == -1) goto werr;
-                /* Remove the loaded object from memory */
-                decrRefCount(po);
-            }
+            expire = getExpire(db,&key);
+            if (rdbSaveKeyValuePair(fp,&key,o,expire,now) == -1) goto werr;

         }
         dictReleaseIterator(di);
     }
         }
         dictReleaseIterator(di);
     }


@@ -507,19 +470,24 @@ werr:
 int rdbSaveBackground(char *filename) {
     pid_t childpid;
 
 int rdbSaveBackground(char *filename) {
     pid_t childpid;
 

-    if (server.bgsavechildpid != -1) return REDIS_ERR;
-    if (server.vm_enabled) waitEmptyIOJobsQueue();
+    if (server.bgsavechildpid != -1 ||
+        server.bgsavethread != (pthread_t) -1) return REDIS_ERR;
+

     server.dirty_before_bgsave = server.dirty;
     server.dirty_before_bgsave = server.dirty;

+
+    if (server.ds_enabled) {
+        cacheForcePointInTime();
+        return dsRdbSaveBackground(filename);
+    }
+

     if ((childpid = fork()) == 0) {
     if ((childpid = fork()) == 0) {

+        int retval;
+

         /* Child */
         /* Child */

-        if (server.vm_enabled) vmReopenSwapFile();

         if (server.ipfd > 0) close(server.ipfd);
         if (server.sofd > 0) close(server.sofd);
         if (server.ipfd > 0) close(server.ipfd);
         if (server.sofd > 0) close(server.sofd);

-        if (rdbSave(filename) == REDIS_OK) {
-            _exit(0);
-        } else {
-            _exit(1);
-        }
+        retval = rdbSave(filename);
+        _exit((retval == REDIS_OK) ? 0 : 1);

     } else {
         /* Parent */
         if (childpid == -1) {
     } else {
         /* Parent */
         if (childpid == -1) {


@@ -839,6 +807,47 @@ robj *rdbLoadObject(int type, FILE *fp) {
                 dictAdd((dict*)o->ptr,key,val);
             }
         }
                 dictAdd((dict*)o->ptr,key,val);
             }
         }

+    } else if (type == REDIS_HASH_ZIPMAP ||
+               type == REDIS_LIST_ZIPLIST ||
+               type == REDIS_SET_INTSET)
+    {
+        robj *aux = rdbLoadStringObject(fp);
+
+        if (aux == NULL) return NULL;
+        o = createObject(REDIS_STRING,NULL); /* string is just placeholder */
+        o->ptr = zmalloc(sdslen(aux->ptr));
+        memcpy(o->ptr,aux->ptr,sdslen(aux->ptr));
+        decrRefCount(aux);
+
+        /* Fix the object encoding, and make sure to convert the encoded
+         * data type into the base type if accordingly to the current
+         * configuration there are too many elements in the encoded data
+         * type. Note that we only check the length and not max element
+         * size as this is an O(N) scan. Eventually everything will get
+         * converted. */
+        switch(type) {
+            case REDIS_HASH_ZIPMAP:
+                o->type = REDIS_HASH;
+                o->encoding = REDIS_ENCODING_ZIPMAP;
+                if (zipmapLen(o->ptr) > server.hash_max_zipmap_entries)
+                    convertToRealHash(o);
+                break;
+            case REDIS_LIST_ZIPLIST:
+                o->type = REDIS_LIST;
+                o->encoding = REDIS_ENCODING_ZIPLIST;
+                if (ziplistLen(o->ptr) > server.list_max_ziplist_entries)
+                    listTypeConvert(o,REDIS_ENCODING_LINKEDLIST);
+                break;
+            case REDIS_SET_INTSET:
+                o->type = REDIS_SET;
+                o->encoding = REDIS_ENCODING_INTSET;
+                if (intsetLen(o->ptr) > server.set_max_intset_entries)
+                    setTypeConvert(o,REDIS_ENCODING_HT);
+                break;
+            default:
+                redisPanic("Unknown enoding");
+                break;
+        }

     } else {
         redisPanic("Unknown object type");
     }
     } else {
         redisPanic("Unknown object type");
     }


@@ -874,7 +883,6 @@ int rdbLoad(char *filename) {
     FILE *fp;
     uint32_t dbid;
     int type, retval, rdbver;
     FILE *fp;
     uint32_t dbid;
     int type, retval, rdbver;

-    int swap_all_values = 0;

     redisDb *db = server.db+0;
     char buf[1024];
     time_t expiretime, now = time(NULL);
     redisDb *db = server.db+0;
     char buf[1024];
     time_t expiretime, now = time(NULL);


@@ -899,8 +907,6 @@ int rdbLoad(char *filename) {
     startLoading(fp);
     while(1) {
         robj *key, *val;
     startLoading(fp);
     while(1) {
         robj *key, *val;

-        int force_swapout;
-

         expiretime = -1;
 
         /* Serve the clients from time to time */
         expiretime = -1;
 
         /* Serve the clients from time to time */


@@ -947,44 +953,7 @@ int rdbLoad(char *filename) {
         /* Set the expire time if needed */
         if (expiretime != -1) setExpire(db,key,expiretime);
 
         /* Set the expire time if needed */
         if (expiretime != -1) setExpire(db,key,expiretime);
 

-        /* Handle swapping while loading big datasets when VM is on */
-
-        /* If we detecter we are hopeless about fitting something in memory
-         * we just swap every new key on disk. Directly...
-         * Note that's important to check for this condition before resorting
-         * to random sampling, otherwise we may try to swap already
-         * swapped keys. */
-        if (swap_all_values) {
-            dictEntry *de = dictFind(db->dict,key->ptr);
-
-            /* de may be NULL since the key already expired */
-            if (de) {
-                vmpointer *vp;
-                val = dictGetEntryVal(de);
-
-                if (val->refcount == 1 &&
-                    (vp = vmSwapObjectBlocking(val)) != NULL)
-                    dictGetEntryVal(de) = vp;
-            }
-            decrRefCount(key);
-            continue;
-        }

         decrRefCount(key);
         decrRefCount(key);

-
-        /* Flush data on disk once 32 MB of additional RAM are used... */
-        force_swapout = 0;
-        if ((zmalloc_used_memory() - server.vm_max_memory) > 1024*1024*32)
-            force_swapout = 1;
-
-        /* If we have still some hope of having some value fitting memory
-         * then we try random sampling. */
-        if (!swap_all_values && server.vm_enabled && force_swapout) {
-            while (zmalloc_used_memory() > server.vm_max_memory) {
-                if (vmSwapOneObjectBlocking() == REDIS_ERR) break;
-            }
-            if (zmalloc_used_memory() > server.vm_max_memory)
-                swap_all_values = 1; /* We are already using too much mem */
-        }

     }
     fclose(fp);
     stopLoading();
     }
     fclose(fp);
     stopLoading();


@@ -997,10 +966,7 @@ eoferr: /* unexpected end of file is handled here with a fatal exit */
 }
 
 /* A background saving child (BGSAVE) terminated its work. Handle this. */
 }
 
 /* A background saving child (BGSAVE) terminated its work. Handle this. */

-void backgroundSaveDoneHandler(int statloc) {
-    int exitcode = WEXITSTATUS(statloc);
-    int bysignal = WIFSIGNALED(statloc);
-
+void backgroundSaveDoneHandler(int exitcode, int bysignal) {

     if (!bysignal && exitcode == 0) {
         redisLog(REDIS_NOTICE,
             "Background saving terminated with success");
     if (!bysignal && exitcode == 0) {
         redisLog(REDIS_NOTICE,
             "Background saving terminated with success");


@@ -1010,11 +976,37 @@ void backgroundSaveDoneHandler(int statloc) {
         redisLog(REDIS_WARNING, "Background saving error");
     } else {
         redisLog(REDIS_WARNING,
         redisLog(REDIS_WARNING, "Background saving error");
     } else {
         redisLog(REDIS_WARNING,

-            "Background saving terminated by signal %d", WTERMSIG(statloc));
+            "Background saving terminated by signal %d", bysignal);

         rdbRemoveTempFile(server.bgsavechildpid);
     }
     server.bgsavechildpid = -1;
         rdbRemoveTempFile(server.bgsavechildpid);
     }
     server.bgsavechildpid = -1;

+    server.bgsavethread = (pthread_t) -1;
+    server.bgsavethread_state = REDIS_BGSAVE_THREAD_UNACTIVE;

     /* Possibly there are slaves waiting for a BGSAVE in order to be served
      * (the first stage of SYNC is a bulk transfer of dump.rdb) */
     updateSlavesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR);
 }
     /* Possibly there are slaves waiting for a BGSAVE in order to be served
      * (the first stage of SYNC is a bulk transfer of dump.rdb) */
     updateSlavesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR);
 }

+
+void saveCommand(redisClient *c) {
+    if (server.bgsavechildpid != -1 || server.bgsavethread != (pthread_t)-1) {
+        addReplyError(c,"Background save already in progress");
+        return;
+    }
+    if (rdbSave(server.dbfilename) == REDIS_OK) {
+        addReply(c,shared.ok);
+    } else {
+        addReply(c,shared.err);
+    }
+}
+
+void bgsaveCommand(redisClient *c) {
+    if (server.bgsavechildpid != -1 || server.bgsavethread != (pthread_t)-1) {
+        addReplyError(c,"Background save already in progress");
+        return;
+    }
+    if (rdbSaveBackground(server.dbfilename) == REDIS_OK) {
+        addReplyStatus(c,"Background saving started");
+    } else {
+        addReply(c,shared.err);
+    }
+}