#include "redis.h"
+#include "bio.h"
#include <signal.h>
#include <fcntl.h>
#include <sys/resource.h>
#include <sys/wait.h>
+void aofUpdateCurrentSize(void);
+
/* Called when the user switches from "appendonly yes" to "appendonly no"
* at runtime using the CONFIG command. */
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;
}
}
* buffer and write it on disk using this function just before entering
* the event loop again. */
void flushAppendOnlyFile(void) {
- time_t now;
ssize_t nwritten;
if (sdslen(server.aofbuf) == 0) return;
* 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);
+
+ /* Perform the fsync if needed. */
if (server.appendfsync == APPENDFSYNC_ALWAYS ||
(server.appendfsync == APPENDFSYNC_EVERYSEC &&
- now-server.lastfsync > 1))
+ server.unixtime > server.lastfsync))
{
/* 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;
}
}
-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) {
c->querybuf = sdsempty();
c->argc = 0;
c->argv = NULL;
+ c->bufpos = 0;
c->flags = 0;
/* We set the fake client as a slave waiting for the synchronization
* so that Redis will not try to send replies to this client. */
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));
server.appendonly = 0;
fakeClient = createFakeClient();
+ startLoading(fp);
+
while(1) {
int argc, j;
unsigned long len;
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))
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;
/* 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.
fclose(fp);
freeFakeClient(fakeClient);
server.appendonly = appendonly;
+ stopLoading();
+ aofUpdateCurrentSize();
+ server.auto_aofrewrite_base_size = server.appendonly_current_size;
return REDIS_OK;
readerr:
}
exit(1);
fmterr:
- redisLog(REDIS_WARNING,"Bad file format reading the append only file");
+ redisLog(REDIS_WARNING,"Bad file format reading the append only file: make a backup of your AOF file, then use ./redis-check-aof --fix <filename>");
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) {
- /* 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);
- } else if (obj->encoding == REDIS_ENCODING_RAW) {
- return fwriteBulkString(fp,obj->ptr,sdslen(obj->ptr));
- } else {
- redisPanic("Unknown string encoding");
- }
-}
-
/* Write a sequence of commands able to fully rebuild the dataset into
* "filename". Used both by REWRITEAOF and BGREWRITEAOF. */
int rewriteAppendOnlyFile(char *filename) {
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;
/* 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;
keystr = dictGetEntryKey(de);
o = dictGetEntryVal(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 */
}
} 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 (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 (vstr != NULL) {
+ if (fwriteBulkString(fp,(char*)vstr,vlen) == 0)
+ goto werr;
+ } else {
+ if (fwriteBulkLongLong(fp,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 = 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;
+ }
+ 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";
if (fwriteBulkObject(fp,&key) == 0) goto werr;
if (fwriteBulkLongLong(fp,expiretime) == 0) goto werr;
}
- if (swapped) decrRefCount(o);
}
dictReleaseIterator(di);
}
*/
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);
}
} else {
/* Parent */
+ server.stat_fork_time = ustime()-start;
if (childpid == -1) {
redisLog(REDIS_WARNING,
"Can't rewrite append only file in background: fork: %s",
void bgrewriteaofCommand(redisClient *c) {
if (server.bgrewritechildpid != -1) {
- addReplySds(c,sdsnew("-ERR background append only file rewriting already in progress\r\n"));
- return;
- }
- if (rewriteAppendOnlyFileBackground() == REDIS_OK) {
- char *status = "+Background append only file rewriting started\r\n";
- addReplySds(c,sdsnew(status));
+ addReplyError(c,"Background append only file rewriting already in progress");
+ } 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);
}
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_NO) aof_fsync(newfd);
+ server.appendseldb = -1; /* Make sure SELECT is re-issued */
+ aofUpdateCurrentSize();
+ server.auto_aofrewrite_base_size = server.appendonly_current_size;
}
+
+ 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();