X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/5171777bf112a6a930c4750949037ce741f3bf28..4be855e757b800a9f848823fcee8b7faf3e1eba0:/src/aof.c

diff --git a/src/aof.c b/src/aof.c
index 16713481..4a463bde 100644
--- a/src/aof.c
+++ b/src/aof.c
@@ -1,4 +1,6 @@
 #include "redis.h"
+#include "bio.h"
+#include "rio.h"
 
 #include <signal.h>
 #include <fcntl.h>
@@ -8,10 +10,16 @@
 #include <sys/resource.h>
 #include <sys/wait.h>
 
+void aofUpdateCurrentSize(void);
+
+void aof_background_fsync(int fd) {
+    bioCreateBackgroundJob(REDIS_BIO_AOF_FSYNC,(void*)(long)fd,NULL,NULL);
+}
+
 /* Called when the user switches from "appendonly yes" to "appendonly no"
  * at runtime using the CONFIG command. */
 void stopAppendOnly(void) {
-    flushAppendOnlyFile();
+    flushAppendOnlyFile(1);
     aof_fsync(server.appendfd);
     close(server.appendfd);
 
@@ -19,15 +27,15 @@ void stopAppendOnly(void) {
     server.appendseldb = -1;
     server.appendonly = 0;
     /* rewrite operation in progress? kill it, wait child exit */
-    if (server.bgsavechildpid != -1) {
+    if (server.bgrewritechildpid != -1) {
         int statloc;
 
-        if (kill(server.bgsavechildpid,SIGKILL) != -1)
+        if (kill(server.bgrewritechildpid,SIGKILL) != -1)
             wait3(&statloc,0,NULL);
         /* reset the buffer accumulating changes while the child saves */
         sdsfree(server.bgrewritebuf);
         server.bgrewritebuf = sdsempty();
-        server.bgsavechildpid = -1;
+        server.bgrewritechildpid = -1;
     }
 }
 
@@ -44,7 +52,7 @@ int startAppendOnly(void) {
     if (rewriteAppendOnlyFileBackground() == REDIS_ERR) {
         server.appendonly = 0;
         close(server.appendfd);
-        redisLog(REDIS_WARNING,"Used tried to switch on AOF via CONFIG, I can't trigger a background AOF rewrite operation. Check the above logs for more info about the error.",strerror(errno));
+        redisLog(REDIS_WARNING,"User tried turning on AOF with CONFIG SET but I can't trigger a background AOF rewrite operation. Check the above logs for more info about the error.");
         return REDIS_ERR;
     }
     return REDIS_OK;
@@ -56,79 +64,155 @@ int startAppendOnly(void) {
  * and the only way the client socket can get a write is entering when the
  * the event loop, we accumulate all the AOF writes in a memory
  * buffer and write it on disk using this function just before entering
- * the event loop again. */
-void flushAppendOnlyFile(void) {
-    time_t now;
+ * the event loop again.
+ *
+ * About the 'force' argument:
+ *
+ * When the fsync policy is set to 'everysec' we may delay the flush if there
+ * is still an fsync() going on in the background thread, since for instance
+ * on Linux write(2) will be blocked by the background fsync anyway.
+ * When this happens we remember that there is some aof buffer to be
+ * flushed ASAP, and will try to do that in the serverCron() function.
+ *
+ * However if force is set to 1 we'll write regardless of the background
+ * fsync. */
+void flushAppendOnlyFile(int force) {
     ssize_t nwritten;
+    int sync_in_progress = 0;
 
     if (sdslen(server.aofbuf) == 0) return;
 
+    if (server.appendfsync == APPENDFSYNC_EVERYSEC)
+        sync_in_progress = bioPendingJobsOfType(REDIS_BIO_AOF_FSYNC) != 0;
+
+    if (server.appendfsync == APPENDFSYNC_EVERYSEC && !force) {
+        /* With this append fsync policy we do background fsyncing.
+         * If the fsync is still in progress we can try to delay
+         * the write for a couple of seconds. */
+        if (sync_in_progress) {
+            if (server.aof_flush_postponed_start == 0) {
+                /* No previous write postponinig, remember that we are
+                 * postponing the flush and return. */
+                server.aof_flush_postponed_start = server.unixtime;
+                return;
+            } else if (server.unixtime - server.aof_flush_postponed_start < 2) {
+                /* We were already waiting for fsync to finish, but for less
+                 * than two seconds this is still ok. Postpone again. */
+                return;
+            }
+            /* Otherwise fall trough, and go write since we can't wait
+             * over two seconds. */
+            redisLog(REDIS_NOTICE,"Asynchronous AOF fsync is taking too long (disk is busy?). Writing the AOF buffer without waiting for fsync to complete, this may slow down Redis.");
+        }
+    }
+    /* If you are following this code path, then we are going to write so
+     * set reset the postponed flush sentinel to zero. */
+    server.aof_flush_postponed_start = 0;
+
     /* We want to perform a single write. This should be guaranteed atomic
      * at least if the filesystem we are writing is a real physical one.
      * While this will save us against the server being killed I don't think
      * there is much to do about the whole server stopping for power problems
      * or alike */
-     nwritten = write(server.appendfd,server.aofbuf,sdslen(server.aofbuf));
-     if (nwritten != (signed)sdslen(server.aofbuf)) {
+    nwritten = write(server.appendfd,server.aofbuf,sdslen(server.aofbuf));
+    if (nwritten != (signed)sdslen(server.aofbuf)) {
         /* Ooops, we are in troubles. The best thing to do for now is
          * aborting instead of giving the illusion that everything is
          * working as expected. */
-         if (nwritten == -1) {
+        if (nwritten == -1) {
             redisLog(REDIS_WARNING,"Exiting on error writing to the append-only file: %s",strerror(errno));
-         } else {
+        } else {
             redisLog(REDIS_WARNING,"Exiting on short write while writing to the append-only file: %s",strerror(errno));
-         }
-         exit(1);
+        }
+        exit(1);
+    }
+    server.appendonly_current_size += nwritten;
+
+    /* Re-use AOF buffer when it is small enough. The maximum comes from the
+     * arena size of 4k minus some overhead (but is otherwise arbitrary). */
+    if ((sdslen(server.aofbuf)+sdsavail(server.aofbuf)) < 4000) {
+        sdsclear(server.aofbuf);
+    } else {
+        sdsfree(server.aofbuf);
+        server.aofbuf = sdsempty();
     }
-    sdsfree(server.aofbuf);
-    server.aofbuf = sdsempty();
 
-    /* Don't Fsync if no-appendfsync-on-rewrite is set to yes and we have
-     * childs performing heavy I/O on disk. */
+    /* Don't fsync if no-appendfsync-on-rewrite is set to yes and there are
+     * children doing I/O in the background. */
     if (server.no_appendfsync_on_rewrite &&
         (server.bgrewritechildpid != -1 || server.bgsavechildpid != -1))
             return;
-    /* Fsync if needed */
-    now = time(NULL);
-    if (server.appendfsync == APPENDFSYNC_ALWAYS ||
-        (server.appendfsync == APPENDFSYNC_EVERYSEC &&
-         now-server.lastfsync > 1))
-    {
+
+    /* Perform the fsync if needed. */
+    if (server.appendfsync == APPENDFSYNC_ALWAYS) {
         /* aof_fsync is defined as fdatasync() for Linux in order to avoid
          * flushing metadata. */
         aof_fsync(server.appendfd); /* Let's try to get this data on the disk */
-        server.lastfsync = now;
+        server.lastfsync = server.unixtime;
+    } else if ((server.appendfsync == APPENDFSYNC_EVERYSEC &&
+                server.unixtime > server.lastfsync)) {
+        if (!sync_in_progress) aof_background_fsync(server.appendfd);
+        server.lastfsync = server.unixtime;
     }
 }
 
-sds catAppendOnlyGenericCommand(sds buf, int argc, robj **argv) {
-    int j;
-    buf = sdscatprintf(buf,"*%d\r\n",argc);
+sds catAppendOnlyGenericCommand(sds dst, int argc, robj **argv) {
+    char buf[32];
+    int len, j;
+    robj *o;
+
+    buf[0] = '*';
+    len = 1+ll2string(buf+1,sizeof(buf)-1,argc);
+    buf[len++] = '\r';
+    buf[len++] = '\n';
+    dst = sdscatlen(dst,buf,len);
+
     for (j = 0; j < argc; j++) {
-        robj *o = getDecodedObject(argv[j]);
-        buf = sdscatprintf(buf,"$%lu\r\n",(unsigned long)sdslen(o->ptr));
-        buf = sdscatlen(buf,o->ptr,sdslen(o->ptr));
-        buf = sdscatlen(buf,"\r\n",2);
+        o = getDecodedObject(argv[j]);
+        buf[0] = '$';
+        len = 1+ll2string(buf+1,sizeof(buf)-1,sdslen(o->ptr));
+        buf[len++] = '\r';
+        buf[len++] = '\n';
+        dst = sdscatlen(dst,buf,len);
+        dst = sdscatlen(dst,o->ptr,sdslen(o->ptr));
+        dst = sdscatlen(dst,"\r\n",2);
         decrRefCount(o);
     }
-    return buf;
+    return dst;
 }
 
-sds catAppendOnlyExpireAtCommand(sds buf, robj *key, robj *seconds) {
-    int argc = 3;
-    long when;
+/* Create the sds representation of an PEXPIREAT command, using
+ * 'seconds' as time to live and 'cmd' to understand what command
+ * we are translating into a PEXPIREAT.
+ *
+ * This command is used in order to translate EXPIRE and PEXPIRE commands
+ * into PEXPIREAT command so that we retain precision in the append only
+ * file, and the time is always absolute and not relative. */
+sds catAppendOnlyExpireAtCommand(sds buf, struct redisCommand *cmd, robj *key, robj *seconds) {
+    long long when;
     robj *argv[3];
 
     /* Make sure we can use strtol */
     seconds = getDecodedObject(seconds);
-    when = time(NULL)+strtol(seconds->ptr,NULL,10);
+    when = strtoll(seconds->ptr,NULL,10);
+    /* Convert argument into milliseconds for EXPIRE, SETEX, EXPIREAT */
+    if (cmd->proc == expireCommand || cmd->proc == setexCommand ||
+        cmd->proc == expireatCommand)
+    {
+        when *= 1000;
+    }
+    /* Convert into absolute time for EXPIRE, PEXPIRE, SETEX, PSETEX */
+    if (cmd->proc == expireCommand || cmd->proc == pexpireCommand ||
+        cmd->proc == setexCommand || cmd->proc == psetexCommand)
+    {
+        when += mstime();
+    }
     decrRefCount(seconds);
 
-    argv[0] = createStringObject("EXPIREAT",8);
+    argv[0] = createStringObject("PEXPIREAT",9);
     argv[1] = key;
-    argv[2] = createObject(REDIS_STRING,
-        sdscatprintf(sdsempty(),"%ld",when));
-    buf = catAppendOnlyGenericCommand(buf, argc, argv);
+    argv[2] = createStringObjectFromLongLong(when);
+    buf = catAppendOnlyGenericCommand(buf, 3, argv);
     decrRefCount(argv[0]);
     decrRefCount(argv[2]);
     return buf;
@@ -149,18 +233,22 @@ void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int a
         server.appendseldb = dictid;
     }
 
-    if (cmd->proc == expireCommand) {
-        /* Translate EXPIRE into EXPIREAT */
-        buf = catAppendOnlyExpireAtCommand(buf,argv[1],argv[2]);
-    } else if (cmd->proc == setexCommand) {
-        /* Translate SETEX to SET and EXPIREAT */
+    if (cmd->proc == expireCommand || cmd->proc == pexpireCommand ||
+        cmd->proc == expireatCommand) {
+        /* Translate EXPIRE/PEXPIRE/EXPIREAT into PEXPIREAT */
+        buf = catAppendOnlyExpireAtCommand(buf,cmd,argv[1],argv[2]);
+    } else if (cmd->proc == setexCommand || cmd->proc == psetexCommand) {
+        /* Translate SETEX/PSETEX to SET and PEXPIREAT */
         tmpargv[0] = createStringObject("SET",3);
         tmpargv[1] = argv[1];
         tmpargv[2] = argv[3];
         buf = catAppendOnlyGenericCommand(buf,3,tmpargv);
         decrRefCount(tmpargv[0]);
-        buf = catAppendOnlyExpireAtCommand(buf,argv[1],argv[2]);
+        buf = catAppendOnlyExpireAtCommand(buf,cmd,argv[1],argv[2]);
     } else {
+        /* All the other commands don't need translation or need the
+         * same translation already operated in the command vector
+         * for the replication itself. */
         buf = catAppendOnlyGenericCommand(buf,argc,argv);
     }
 
@@ -218,9 +306,13 @@ int loadAppendOnlyFile(char *filename) {
     FILE *fp = fopen(filename,"r");
     struct redis_stat sb;
     int appendonly = server.appendonly;
+    long loops = 0;
 
-    if (redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0)
+    if (fp && redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0) {
+        server.appendonly_current_size = 0;
+        fclose(fp);
         return REDIS_ERR;
+    }
 
     if (fp == NULL) {
         redisLog(REDIS_WARNING,"Fatal error: can't open the append log file for reading: %s",strerror(errno));
@@ -232,6 +324,8 @@ int loadAppendOnlyFile(char *filename) {
     server.appendonly = 0;
 
     fakeClient = createFakeClient();
+    startLoading(fp);
+
     while(1) {
         int argc, j;
         unsigned long len;
@@ -239,7 +333,12 @@ int loadAppendOnlyFile(char *filename) {
         char buf[128];
         sds argsds;
         struct redisCommand *cmd;
-        int force_swapout;
+
+        /* Serve the clients from time to time */
+        if (!(loops++ % 1000)) {
+            loadingProgress(ftello(fp));
+            aeProcessEvents(server.el, AE_FILE_EVENTS|AE_DONT_WAIT);
+        }
 
         if (fgets(buf,sizeof(buf),fp) == NULL) {
             if (feof(fp))
@@ -249,6 +348,8 @@ int loadAppendOnlyFile(char *filename) {
         }
         if (buf[0] != '*') goto fmterr;
         argc = atoi(buf+1);
+        if (argc < 1) goto fmterr;
+
         argv = zmalloc(sizeof(robj*)*argc);
         for (j = 0; j < argc; j++) {
             if (fgets(buf,sizeof(buf),fp) == NULL) goto readerr;
@@ -266,9 +367,6 @@ int loadAppendOnlyFile(char *filename) {
             redisLog(REDIS_WARNING,"Unknown command '%s' reading the append only file", argv[0]->ptr);
             exit(1);
         }
-        /* Try object encoding */
-        if (cmd->flags & REDIS_CMD_BULK)
-            argv[argc-1] = tryObjectEncoding(argv[argc-1]);
         /* Run the command in the context of a fake client */
         fakeClient->argc = argc;
         fakeClient->argv = argv;
@@ -276,21 +374,14 @@ int loadAppendOnlyFile(char *filename) {
 
         /* The fake client should not have a reply */
         redisAssert(fakeClient->bufpos == 0 && listLength(fakeClient->reply) == 0);
-
-        /* Clean up, ready for the next command */
-        for (j = 0; j < argc; j++) decrRefCount(argv[j]);
-        zfree(argv);
-
-        /* Handle swapping while loading big datasets when VM is on */
-        force_swapout = 0;
-        if ((zmalloc_used_memory() - server.vm_max_memory) > 1024*1024*32)
-            force_swapout = 1;
-
-        if (server.vm_enabled && force_swapout) {
-            while (zmalloc_used_memory() > server.vm_max_memory) {
-                if (vmSwapOneObjectBlocking() == REDIS_ERR) break;
-            }
-        }
+        /* The fake client should never get blocked */
+        redisAssert((fakeClient->flags & REDIS_BLOCKED) == 0);
+
+        /* Clean up. Command code may have changed argv/argc so we use the
+         * argv/argc of the client instead of the local variables. */
+        for (j = 0; j < fakeClient->argc; j++)
+            decrRefCount(fakeClient->argv[j]);
+        zfree(fakeClient->argv);
     }
 
     /* This point can only be reached when EOF is reached without errors.
@@ -300,6 +391,9 @@ int loadAppendOnlyFile(char *filename) {
     fclose(fp);
     freeFakeClient(fakeClient);
     server.appendonly = appendonly;
+    stopLoading();
+    aofUpdateCurrentSize();
+    server.auto_aofrewrite_base_size = server.appendonly_current_size;
     return REDIS_OK;
 
 readerr:
@@ -314,50 +408,15 @@ fmterr:
     exit(1);
 }
 
-/* Write binary-safe string into a file in the bulkformat
- * $<count>\r\n<payload>\r\n */
-int fwriteBulkString(FILE *fp, char *s, unsigned long len) {
-    char cbuf[128];
-    int clen;
-    cbuf[0] = '$';
-    clen = 1+ll2string(cbuf+1,sizeof(cbuf)-1,len);
-    cbuf[clen++] = '\r';
-    cbuf[clen++] = '\n';
-    if (fwrite(cbuf,clen,1,fp) == 0) return 0;
-    if (len > 0 && fwrite(s,len,1,fp) == 0) return 0;
-    if (fwrite("\r\n",2,1,fp) == 0) return 0;
-    return 1;
-}
-
-/* Write a double value in bulk format $<count>\r\n<payload>\r\n */
-int fwriteBulkDouble(FILE *fp, double d) {
-    char buf[128], dbuf[128];
-
-    snprintf(dbuf,sizeof(dbuf),"%.17g\r\n",d);
-    snprintf(buf,sizeof(buf),"$%lu\r\n",(unsigned long)strlen(dbuf)-2);
-    if (fwrite(buf,strlen(buf),1,fp) == 0) return 0;
-    if (fwrite(dbuf,strlen(dbuf),1,fp) == 0) return 0;
-    return 1;
-}
-
-/* Write a long value in bulk format $<count>\r\n<payload>\r\n */
-int fwriteBulkLongLong(FILE *fp, long long l) {
-    char bbuf[128], lbuf[128];
-    unsigned int blen, llen;
-    llen = ll2string(lbuf,32,l);
-    blen = snprintf(bbuf,sizeof(bbuf),"$%u\r\n%s\r\n",llen,lbuf);
-    if (fwrite(bbuf,blen,1,fp) == 0) return 0;
-    return 1;
-}
-
-/* Delegate writing an object to writing a bulk string or bulk long long. */
-int fwriteBulkObject(FILE *fp, robj *obj) {
+/* Delegate writing an object to writing a bulk string or bulk long long.
+ * This is not placed in rio.c since that adds the redis.h dependency. */
+int rioWriteBulkObject(rio *r, robj *obj) {
     /* Avoid using getDecodedObject to help copy-on-write (we are often
      * in a child process when this function is called). */
     if (obj->encoding == REDIS_ENCODING_INT) {
-        return fwriteBulkLongLong(fp,(long)obj->ptr);
+        return rioWriteBulkLongLong(r,(long)obj->ptr);
     } else if (obj->encoding == REDIS_ENCODING_RAW) {
-        return fwriteBulkString(fp,obj->ptr,sdslen(obj->ptr));
+        return rioWriteBulkString(r,obj->ptr,sdslen(obj->ptr));
     } else {
         redisPanic("Unknown string encoding");
     }
@@ -368,10 +427,11 @@ int fwriteBulkObject(FILE *fp, robj *obj) {
 int rewriteAppendOnlyFile(char *filename) {
     dictIterator *di = NULL;
     dictEntry *de;
+    rio aof;
     FILE *fp;
     char tmpfile[256];
     int j;
-    time_t now = time(NULL);
+    long long now = mstime();
 
     /* Note that we have to use a different temp name here compared to the
      * one used by rewriteAppendOnlyFileBackground() function. */
@@ -381,52 +441,43 @@ int rewriteAppendOnlyFile(char *filename) {
         redisLog(REDIS_WARNING, "Failed rewriting the append only file: %s", strerror(errno));
         return REDIS_ERR;
     }
+
+    rioInitWithFile(&aof,fp);
     for (j = 0; j < server.dbnum; j++) {
         char selectcmd[] = "*2\r\n$6\r\nSELECT\r\n";
         redisDb *db = server.db+j;
         dict *d = db->dict;
         if (dictSize(d) == 0) continue;
-        di = dictGetIterator(d);
+        di = dictGetSafeIterator(d);
         if (!di) {
             fclose(fp);
             return REDIS_ERR;
         }
 
         /* SELECT the new DB */
-        if (fwrite(selectcmd,sizeof(selectcmd)-1,1,fp) == 0) goto werr;
-        if (fwriteBulkLongLong(fp,j) == 0) goto werr;
+        if (rioWrite(&aof,selectcmd,sizeof(selectcmd)-1) == 0) goto werr;
+        if (rioWriteBulkLongLong(&aof,j) == 0) goto werr;
 
         /* Iterate this DB writing every entry */
         while((de = dictNext(di)) != NULL) {
-            sds keystr = dictGetEntryKey(de);
+            sds keystr;
             robj key, *o;
-            time_t expiretime;
-            int swapped;
+            long long expiretime;
 
-            keystr = dictGetEntryKey(de);
-            o = dictGetEntryVal(de);
+            keystr = dictGetKey(de);
+            o = dictGetVal(de);
             initStaticStringObject(key,keystr);
-            /* If the value for this key is swapped, load a preview in memory.
-             * We use a "swapped" flag to remember if we need to free the
-             * value object instead to just increment the ref count anyway
-             * in order to avoid copy-on-write of pages if we are forked() */
-            if (!server.vm_enabled || o->storage == REDIS_VM_MEMORY ||
-                o->storage == REDIS_VM_SWAPPING) {
-                swapped = 0;
-            } else {
-                o = vmPreviewObject(o);
-                swapped = 1;
-            }
+
             expiretime = getExpire(db,&key);
 
             /* Save the key and associated value */
             if (o->type == REDIS_STRING) {
                 /* Emit a SET command */
                 char cmd[]="*3\r\n$3\r\nSET\r\n";
-                if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+                if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;
                 /* Key and value */
-                if (fwriteBulkObject(fp,&key) == 0) goto werr;
-                if (fwriteBulkObject(fp,o) == 0) goto werr;
+                if (rioWriteBulkObject(&aof,&key) == 0) goto werr;
+                if (rioWriteBulkObject(&aof,o) == 0) goto werr;
             } else if (o->type == REDIS_LIST) {
                 /* Emit the RPUSHes needed to rebuild the list */
                 char cmd[]="*3\r\n$5\r\nRPUSH\r\n";
@@ -438,13 +489,13 @@ int rewriteAppendOnlyFile(char *filename) {
                     long long vlong;
 
                     while(ziplistGet(p,&vstr,&vlen,&vlong)) {
-                        if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
-                        if (fwriteBulkObject(fp,&key) == 0) goto werr;
+                        if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,&key) == 0) goto werr;
                         if (vstr) {
-                            if (fwriteBulkString(fp,(char*)vstr,vlen) == 0)
+                            if (rioWriteBulkString(&aof,(char*)vstr,vlen) == 0)
                                 goto werr;
                         } else {
-                            if (fwriteBulkLongLong(fp,vlong) == 0)
+                            if (rioWriteBulkLongLong(&aof,vlong) == 0)
                                 goto werr;
                         }
                         p = ziplistNext(zl,p);
@@ -458,9 +509,9 @@ int rewriteAppendOnlyFile(char *filename) {
                     while((ln = listNext(&li))) {
                         robj *eleobj = listNodeValue(ln);
 
-                        if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
-                        if (fwriteBulkObject(fp,&key) == 0) goto werr;
-                        if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
+                        if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,&key) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,eleobj) == 0) goto werr;
                     }
                 } else {
                     redisPanic("Unknown list encoding");
@@ -473,18 +524,18 @@ int rewriteAppendOnlyFile(char *filename) {
                     int ii = 0;
                     int64_t llval;
                     while(intsetGet(o->ptr,ii++,&llval)) {
-                        if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
-                        if (fwriteBulkObject(fp,&key) == 0) goto werr;
-                        if (fwriteBulkLongLong(fp,llval) == 0) goto werr;
+                        if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,&key) == 0) goto werr;
+                        if (rioWriteBulkLongLong(&aof,llval) == 0) goto werr;
                     }
                 } else if (o->encoding == REDIS_ENCODING_HT) {
                     dictIterator *di = dictGetIterator(o->ptr);
                     dictEntry *de;
                     while((de = dictNext(di)) != NULL) {
-                        robj *eleobj = dictGetEntryKey(de);
-                        if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
-                        if (fwriteBulkObject(fp,&key) == 0) goto werr;
-                        if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
+                        robj *eleobj = dictGetKey(de);
+                        if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,&key) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,eleobj) == 0) goto werr;
                     }
                     dictReleaseIterator(di);
                 } else {
@@ -492,21 +543,55 @@ int rewriteAppendOnlyFile(char *filename) {
                 }
             } else if (o->type == REDIS_ZSET) {
                 /* Emit the ZADDs needed to rebuild the sorted set */
-                zset *zs = o->ptr;
-                dictIterator *di = dictGetIterator(zs->dict);
-                dictEntry *de;
-
-                while((de = dictNext(di)) != NULL) {
-                    char cmd[]="*4\r\n$4\r\nZADD\r\n";
-                    robj *eleobj = dictGetEntryKey(de);
-                    double *score = dictGetEntryVal(de);
-
-                    if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
-                    if (fwriteBulkObject(fp,&key) == 0) goto werr;
-                    if (fwriteBulkDouble(fp,*score) == 0) goto werr;
-                    if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
+                char cmd[]="*4\r\n$4\r\nZADD\r\n";
+
+                if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+                    unsigned char *zl = o->ptr;
+                    unsigned char *eptr, *sptr;
+                    unsigned char *vstr;
+                    unsigned int vlen;
+                    long long vll;
+                    double score;
+
+                    eptr = ziplistIndex(zl,0);
+                    redisAssert(eptr != NULL);
+                    sptr = ziplistNext(zl,eptr);
+                    redisAssert(sptr != NULL);
+
+                    while (eptr != NULL) {
+                        redisAssert(ziplistGet(eptr,&vstr,&vlen,&vll));
+                        score = zzlGetScore(sptr);
+
+                        if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,&key) == 0) goto werr;
+                        if (rioWriteBulkDouble(&aof,score) == 0) goto werr;
+                        if (vstr != NULL) {
+                            if (rioWriteBulkString(&aof,(char*)vstr,vlen) == 0)
+                                goto werr;
+                        } else {
+                            if (rioWriteBulkLongLong(&aof,vll) == 0)
+                                goto werr;
+                        }
+                        zzlNext(zl,&eptr,&sptr);
+                    }
+                } else if (o->encoding == REDIS_ENCODING_SKIPLIST) {
+                    zset *zs = o->ptr;
+                    dictIterator *di = dictGetIterator(zs->dict);
+                    dictEntry *de;
+
+                    while((de = dictNext(di)) != NULL) {
+                        robj *eleobj = dictGetKey(de);
+                        double *score = dictGetVal(de);
+
+                        if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,&key) == 0) goto werr;
+                        if (rioWriteBulkDouble(&aof,*score) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,eleobj) == 0) goto werr;
+                    }
+                    dictReleaseIterator(di);
+                } else {
+                    redisPanic("Unknown sorted set encoding");
                 }
-                dictReleaseIterator(di);
             } else if (o->type == REDIS_HASH) {
                 char cmd[]="*4\r\n$4\r\nHSET\r\n";
 
@@ -517,11 +602,11 @@ int rewriteAppendOnlyFile(char *filename) {
                     unsigned int flen, vlen;
 
                     while((p = zipmapNext(p,&field,&flen,&val,&vlen)) != NULL) {
-                        if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
-                        if (fwriteBulkObject(fp,&key) == 0) goto werr;
-                        if (fwriteBulkString(fp,(char*)field,flen) == 0)
+                        if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,&key) == 0) goto werr;
+                        if (rioWriteBulkString(&aof,(char*)field,flen) == 0)
                             goto werr;
-                        if (fwriteBulkString(fp,(char*)val,vlen) == 0)
+                        if (rioWriteBulkString(&aof,(char*)val,vlen) == 0)
                             goto werr;
                     }
                 } else {
@@ -529,13 +614,13 @@ int rewriteAppendOnlyFile(char *filename) {
                     dictEntry *de;
 
                     while((de = dictNext(di)) != NULL) {
-                        robj *field = dictGetEntryKey(de);
-                        robj *val = dictGetEntryVal(de);
+                        robj *field = dictGetKey(de);
+                        robj *val = dictGetVal(de);
 
-                        if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
-                        if (fwriteBulkObject(fp,&key) == 0) goto werr;
-                        if (fwriteBulkObject(fp,field) == 0) goto werr;
-                        if (fwriteBulkObject(fp,val) == 0) goto werr;
+                        if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,&key) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,field) == 0) goto werr;
+                        if (rioWriteBulkObject(&aof,val) == 0) goto werr;
                     }
                     dictReleaseIterator(di);
                 }
@@ -544,14 +629,13 @@ int rewriteAppendOnlyFile(char *filename) {
             }
             /* Save the expire time */
             if (expiretime != -1) {
-                char cmd[]="*3\r\n$8\r\nEXPIREAT\r\n";
+                char cmd[]="*3\r\n$9\r\nPEXPIREAT\r\n";
                 /* If this key is already expired skip it */
                 if (expiretime < now) continue;
-                if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
-                if (fwriteBulkObject(fp,&key) == 0) goto werr;
-                if (fwriteBulkLongLong(fp,expiretime) == 0) goto werr;
+                if (rioWrite(&aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                if (rioWriteBulkObject(&aof,&key) == 0) goto werr;
+                if (rioWriteBulkLongLong(&aof,expiretime) == 0) goto werr;
             }
-            if (swapped) decrRefCount(o);
         }
         dictReleaseIterator(di);
     }
@@ -593,15 +677,16 @@ werr:
  */
 int rewriteAppendOnlyFileBackground(void) {
     pid_t childpid;
+    long long start;
 
     if (server.bgrewritechildpid != -1) return REDIS_ERR;
-    if (server.vm_enabled) waitEmptyIOJobsQueue();
+    start = ustime();
     if ((childpid = fork()) == 0) {
-        /* Child */
         char tmpfile[256];
 
-        if (server.vm_enabled) vmReopenSwapFile();
-        close(server.fd);
+        /* Child */
+        if (server.ipfd > 0) close(server.ipfd);
+        if (server.sofd > 0) close(server.sofd);
         snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) getpid());
         if (rewriteAppendOnlyFile(tmpfile) == REDIS_OK) {
             _exit(0);
@@ -610,6 +695,7 @@ int rewriteAppendOnlyFileBackground(void) {
         }
     } else {
         /* Parent */
+        server.stat_fork_time = ustime()-start;
         if (childpid == -1) {
             redisLog(REDIS_WARNING,
                 "Can't rewrite append only file in background: fork: %s",
@@ -618,6 +704,7 @@ int rewriteAppendOnlyFileBackground(void) {
         }
         redisLog(REDIS_NOTICE,
             "Background append only file rewriting started by pid %d",childpid);
+        server.aofrewrite_scheduled = 0;
         server.bgrewritechildpid = childpid;
         updateDictResizePolicy();
         /* We set appendseldb to -1 in order to force the next call to the
@@ -633,9 +720,10 @@ int rewriteAppendOnlyFileBackground(void) {
 void bgrewriteaofCommand(redisClient *c) {
     if (server.bgrewritechildpid != -1) {
         addReplyError(c,"Background append only file rewriting already in progress");
-        return;
-    }
-    if (rewriteAppendOnlyFileBackground() == REDIS_OK) {
+    } else if (server.bgsavechildpid != -1) {
+        server.aofrewrite_scheduled = 1;
+        addReplyStatus(c,"Background append only file rewriting scheduled");
+    } else if (rewriteAppendOnlyFileBackground() == REDIS_OK) {
         addReplyStatus(c,"Background append only file rewriting started");
     } else {
         addReply(c,shared.err);
@@ -649,61 +737,146 @@ void aofRemoveTempFile(pid_t childpid) {
     unlink(tmpfile);
 }
 
+/* Update the server.appendonly_current_size filed explicitly using stat(2)
+ * to check the size of the file. This is useful after a rewrite or after
+ * a restart, normally the size is updated just adding the write length
+ * to the current lenght, that is much faster. */
+void aofUpdateCurrentSize(void) {
+    struct redis_stat sb;
+
+    if (redis_fstat(server.appendfd,&sb) == -1) {
+        redisLog(REDIS_WARNING,"Unable to check the AOF length: %s",
+            strerror(errno));
+    } else {
+        server.appendonly_current_size = sb.st_size;
+    }
+}
+
 /* A background append only file rewriting (BGREWRITEAOF) terminated its work.
  * Handle this. */
-void backgroundRewriteDoneHandler(int statloc) {
-    int exitcode = WEXITSTATUS(statloc);
-    int bysignal = WIFSIGNALED(statloc);
-
+void backgroundRewriteDoneHandler(int exitcode, int bysignal) {
     if (!bysignal && exitcode == 0) {
-        int fd;
+        int newfd, oldfd;
+        int nwritten;
         char tmpfile[256];
+        long long now = ustime();
 
         redisLog(REDIS_NOTICE,
-            "Background append only file rewriting terminated with success");
-        /* Now it's time to flush the differences accumulated by the parent */
-        snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) server.bgrewritechildpid);
-        fd = open(tmpfile,O_WRONLY|O_APPEND);
-        if (fd == -1) {
-            redisLog(REDIS_WARNING, "Not able to open the temp append only file produced by the child: %s", strerror(errno));
+            "Background AOF rewrite terminated with success");
+
+        /* Flush the differences accumulated by the parent to the
+         * rewritten AOF. */
+        snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof",
+            (int)server.bgrewritechildpid);
+        newfd = open(tmpfile,O_WRONLY|O_APPEND);
+        if (newfd == -1) {
+            redisLog(REDIS_WARNING,
+                "Unable to open the temporary AOF produced by the child: %s", strerror(errno));
             goto cleanup;
         }
-        /* Flush our data... */
-        if (write(fd,server.bgrewritebuf,sdslen(server.bgrewritebuf)) !=
-                (signed) sdslen(server.bgrewritebuf)) {
-            redisLog(REDIS_WARNING, "Error or short write trying to flush the parent diff of the append log file in the child temp file: %s", strerror(errno));
-            close(fd);
+
+        nwritten = write(newfd,server.bgrewritebuf,sdslen(server.bgrewritebuf));
+        if (nwritten != (signed)sdslen(server.bgrewritebuf)) {
+            if (nwritten == -1) {
+                redisLog(REDIS_WARNING,
+                    "Error trying to flush the parent diff to the rewritten AOF: %s", strerror(errno));
+            } else {
+                redisLog(REDIS_WARNING,
+                    "Short write trying to flush the parent diff to the rewritten AOF: %s", strerror(errno));
+            }
+            close(newfd);
             goto cleanup;
         }
-        redisLog(REDIS_NOTICE,"Parent diff flushed into the new append log file with success (%lu bytes)",sdslen(server.bgrewritebuf));
-        /* Now our work is to rename the temp file into the stable file. And
-         * switch the file descriptor used by the server for append only. */
+
+        redisLog(REDIS_NOTICE,
+            "Parent diff successfully flushed to the rewritten AOF (%lu bytes)", nwritten);
+
+        /* The only remaining thing to do is to rename the temporary file to
+         * the configured file and switch the file descriptor used to do AOF
+         * writes. We don't want close(2) or rename(2) calls to block the
+         * server on old file deletion.
+         *
+         * There are two possible scenarios:
+         *
+         * 1) AOF is DISABLED and this was a one time rewrite. The temporary
+         * file will be renamed to the configured file. When this file already
+         * exists, it will be unlinked, which may block the server.
+         *
+         * 2) AOF is ENABLED and the rewritten AOF will immediately start
+         * receiving writes. After the temporary file is renamed to the
+         * configured file, the original AOF file descriptor will be closed.
+         * Since this will be the last reference to that file, closing it
+         * causes the underlying file to be unlinked, which may block the
+         * server.
+         *
+         * To mitigate the blocking effect of the unlink operation (either
+         * caused by rename(2) in scenario 1, or by close(2) in scenario 2), we
+         * use a background thread to take care of this. First, we
+         * make scenario 1 identical to scenario 2 by opening the target file
+         * when it exists. The unlink operation after the rename(2) will then
+         * be executed upon calling close(2) for its descriptor. Everything to
+         * guarantee atomicity for this switch has already happened by then, so
+         * we don't care what the outcome or duration of that close operation
+         * is, as long as the file descriptor is released again. */
+        if (server.appendfd == -1) {
+            /* AOF disabled */
+
+             /* Don't care if this fails: oldfd will be -1 and we handle that.
+              * One notable case of -1 return is if the old file does
+              * not exist. */
+             oldfd = open(server.appendfilename,O_RDONLY|O_NONBLOCK);
+        } else {
+            /* AOF enabled */
+            oldfd = -1; /* We'll set this to the current AOF filedes later. */
+        }
+
+        /* Rename the temporary file. This will not unlink the target file if
+         * it exists, because we reference it with "oldfd". */
         if (rename(tmpfile,server.appendfilename) == -1) {
-            redisLog(REDIS_WARNING,"Can't rename the temp append only file into the stable one: %s", strerror(errno));
-            close(fd);
+            redisLog(REDIS_WARNING,
+                "Error trying to rename the temporary AOF: %s", strerror(errno));
+            close(newfd);
+            if (oldfd != -1) close(oldfd);
             goto cleanup;
         }
-        /* Mission completed... almost */
-        redisLog(REDIS_NOTICE,"Append only file successfully rewritten.");
-        if (server.appendfd != -1) {
-            /* If append only is actually enabled... */
-            close(server.appendfd);
-            server.appendfd = fd;
-            if (server.appendfsync != APPENDFSYNC_NO) aof_fsync(fd);
-            server.appendseldb = -1; /* Make sure it will issue SELECT */
-            redisLog(REDIS_NOTICE,"The new append only file was selected for future appends.");
+
+        if (server.appendfd == -1) {
+            /* AOF disabled, we don't need to set the AOF file descriptor
+             * to this new file, so we can close it. */
+            close(newfd);
         } else {
-            /* If append only is disabled we just generate a dump in this
-             * format. Why not? */
-            close(fd);
+            /* AOF enabled, replace the old fd with the new one. */
+            oldfd = server.appendfd;
+            server.appendfd = newfd;
+            if (server.appendfsync == APPENDFSYNC_ALWAYS)
+                aof_fsync(newfd);
+            else if (server.appendfsync == APPENDFSYNC_EVERYSEC)
+                aof_background_fsync(newfd);
+            server.appendseldb = -1; /* Make sure SELECT is re-issued */
+            aofUpdateCurrentSize();
+            server.auto_aofrewrite_base_size = server.appendonly_current_size;
+
+            /* Clear regular AOF buffer since its contents was just written to
+             * the new AOF from the background rewrite buffer. */
+            sdsfree(server.aofbuf);
+            server.aofbuf = sdsempty();
         }
+
+        redisLog(REDIS_NOTICE, "Background AOF rewrite successful");
+
+        /* Asynchronously close the overwritten AOF. */
+        if (oldfd != -1) bioCreateBackgroundJob(REDIS_BIO_CLOSE_FILE,(void*)(long)oldfd,NULL,NULL);
+
+        redisLog(REDIS_VERBOSE,
+            "Background AOF rewrite signal handler took %lldus", ustime()-now);
     } else if (!bysignal && exitcode != 0) {
-        redisLog(REDIS_WARNING, "Background append only file rewriting error");
+        redisLog(REDIS_WARNING,
+            "Background AOF rewrite terminated with error");
     } else {
         redisLog(REDIS_WARNING,
-            "Background append only file rewriting terminated by signal %d",
-            WTERMSIG(statloc));
+            "Background AOF rewrite terminated by signal %d", bysignal);
     }
+
 cleanup:
     sdsfree(server.bgrewritebuf);
     server.bgrewritebuf = sdsempty();