int daemonize;
int appendonly;
int appendfsync;
+ int no_appendfsync_on_rewrite;
int shutdown_asap;
time_t lastfsync;
int appendfd;
/* If append only is actually enabled... */
close(server.appendfd);
server.appendfd = fd;
- fsync(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.");
} else {
server.daemonize = 0;
server.appendonly = 0;
server.appendfsync = APPENDFSYNC_EVERYSEC;
+ server.no_appendfsync_on_rewrite = 0;
server.lastfsync = time(NULL);
server.appendfd = -1;
server.appendseldb = -1; /* Make sure the first time will not match */
} else if (!strcasecmp(argv[0],"appendfilename") && argc == 2) {
zfree(server.appendfilename);
server.appendfilename = zstrdup(argv[1]);
+ } else if (!strcasecmp(argv[0],"no-appendfsync-on-rewrite")
+ && argc == 2) {
+ if ((server.no_appendfsync_on_rewrite= yesnotoi(argv[1])) == -1) {
+ err = "argument must be 'yes' or 'no'"; goto loaderr;
+ }
} else if (!strcasecmp(argv[0],"appendfsync") && argc == 2) {
if (!strcasecmp(argv[1],"no")) {
server.appendfsync = APPENDFSYNC_NO;
redisDb *db = server.db+0;
char buf[1024];
time_t expiretime, now = time(NULL);
- long long loadedkeys = 0;
fp = fopen(filename,"r");
if (!fp) return REDIS_ERR;
}
while(1) {
robj *key, *val;
+ int force_swapout;
expiretime = -1;
/* Read type. */
redisLog(REDIS_WARNING,"Loading DB, duplicated key (%s) found! Unrecoverable error, exiting now.", key->ptr);
exit(1);
}
- loadedkeys++;
/* Set the expire time if needed */
if (expiretime != -1) setExpire(db,key,expiretime);
continue;
}
+ /* 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 && (loadedkeys % 5000) == 0) {
+ if (!swap_all_values && server.vm_enabled && force_swapout) {
while (zmalloc_used_memory() > server.vm_max_memory) {
if (vmSwapOneObjectBlocking() == REDIS_ERR) break;
}
}
if (server.appendonly) {
/* Append only file: fsync() the AOF and exit */
- fsync(server.appendfd);
+ aof_fsync(server.appendfd);
if (server.vm_enabled) unlink(server.vm_swap_file);
} else {
/* Snapshotting. Perform a SYNC SAVE and exit */
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. */
+ if (server.no_appendfsync_on_rewrite &&
+ (server.bgrewritechildpid != -1 || server.bgsavechildpid != -1))
+ return;
/* Fsync if needed */
now = time(NULL);
if (server.appendfsync == APPENDFSYNC_ALWAYS ||
struct redisClient *fakeClient;
FILE *fp = fopen(filename,"r");
struct redis_stat sb;
- unsigned long long loadedkeys = 0;
int appendonly = server.appendonly;
if (redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0)
char buf[128];
sds argsds;
struct redisCommand *cmd;
+ int force_swapout;
if (fgets(buf,sizeof(buf),fp) == NULL) {
if (feof(fp))
for (j = 0; j < argc; j++) decrRefCount(argv[j]);
zfree(argv);
/* Handle swapping while loading big datasets when VM is on */
- loadedkeys++;
- if (server.vm_enabled && (loadedkeys % 5000) == 0) {
+ 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;
}
/* Make sure data will not remain on the OS's output buffers */
fflush(fp);
- fsync(fileno(fp));
+ aof_fsync(fileno(fp));
fclose(fp);
/* Use RENAME to make sure the DB file is changed atomically only
* at runtime using the CONFIG command. */
static void stopAppendOnly(void) {
flushAppendOnlyFile();
- fsync(server.appendfd);
+ aof_fsync(server.appendfd);
close(server.appendfd);
server.appendfd = -1;
} else {
goto badfmt;
}
+ } else if (!strcasecmp(c->argv[2]->ptr,"no-appendfsync-on-rewrite")) {
+ int yn = yesnotoi(o->ptr);
+
+ if (yn == -1) goto badfmt;
+ server.no_appendfsync_on_rewrite = yn;
} else if (!strcasecmp(c->argv[2]->ptr,"appendonly")) {
int old = server.appendonly;
int new = yesnotoi(o->ptr);
addReplyBulkCString(c,server.appendonly ? "yes" : "no");
matches++;
}
+ if (stringmatch(pattern,"no-appendfsync-on-rewrite",0)) {
+ addReplyBulkCString(c,"no-appendfsync-on-rewrite");
+ addReplyBulkCString(c,server.no_appendfsync_on_rewrite ? "yes" : "no");
+ matches++;
+ }
if (stringmatch(pattern,"appendfsync",0)) {
char *policy;
appendfsync everysec
# appendfsync no
+# When the AOF fsync policy is set to always or everysec, and a background
+# saving process (a background save or AOF log background rewriting) is
+# performing a lot of I/O against the disk, in some Linux configurations
+# Redis may block too long on the fsync() call. Note that there is no fix for
+# this currently, as even performing fsync in a different thread will block
+# our synchronous write(2) call.
+#
+# In order to mitigate this problem it's possible to use the following option
+# that will prevent fsync() from being called in the main process while a
+# BGSAVE or BGREWRITEAOF is in progress.
+#
+# This means that while another child is saving the durability of Redis is
+# the same as "appendfsync none", that in pratical terms means that it is
+# possible to lost up to 30 seconds of log in the worst scenario (with the
+# default Linux settings).
+#
+# If you have latency problems turn this to "yes". Otherwise leave it as
+# "no" that is the safest pick from the point of view of durability.
+no-appendfsync-on-rewrite no
+
################################ VIRTUAL MEMORY ###############################
# Virtual Memory allows Redis to work with datasets bigger than the actual