config option to select if when replication link with master a slave should or not...

[redis.git] / src / redis.h

diff --git a/src/redis.h b/src/redis.h
index 9e27d724854e1788199286f83e9437c81b7df954..043825af769708981063a726613425d5842a10f4 100644 (file)
--- a/src/redis.h
+++ b/src/redis.h
@@ -41,12 +41,12 @@
 #define REDIS_STATIC_ARGS       8
 #define REDIS_DEFAULT_DBNUM     16
 #define REDIS_CONFIGLINE_MAX    1024
 #define REDIS_STATIC_ARGS       8
 #define REDIS_DEFAULT_DBNUM     16
 #define REDIS_CONFIGLINE_MAX    1024

-#define REDIS_OBJFREELIST_MAX   1000000 /* Max number of objects to cache */

 #define REDIS_MAX_SYNC_TIME     60      /* Slave can't take more to sync */
 #define REDIS_EXPIRELOOKUPS_PER_CRON    10 /* lookup 10 expires per loop */
 #define REDIS_MAX_WRITE_PER_EVENT (1024*64)
 #define REDIS_REQUEST_MAX_SIZE (1024*1024*256) /* max bytes in inline command */
 #define REDIS_SHARED_INTEGERS 10000
 #define REDIS_MAX_SYNC_TIME     60      /* Slave can't take more to sync */
 #define REDIS_EXPIRELOOKUPS_PER_CRON    10 /* lookup 10 expires per loop */
 #define REDIS_MAX_WRITE_PER_EVENT (1024*64)
 #define REDIS_REQUEST_MAX_SIZE (1024*1024*256) /* max bytes in inline command */
 #define REDIS_SHARED_INTEGERS 10000

+#define REDIS_REPLY_CHUNK_BYTES (5*1500) /* 5 TCP packets with default MTU */

 
 /* If more then REDIS_WRITEV_THRESHOLD write packets are pending use writev */
 #define REDIS_WRITEV_THRESHOLD      3
 
 /* If more then REDIS_WRITEV_THRESHOLD write packets are pending use writev */
 #define REDIS_WRITEV_THRESHOLD      3


@@ -56,15 +56,15 @@
 /* Hash table parameters */
 #define REDIS_HT_MINFILL        10      /* Minimal hash table fill 10% */
 
 /* Hash table parameters */
 #define REDIS_HT_MINFILL        10      /* Minimal hash table fill 10% */
 

-/* Command flags */
-#define REDIS_CMD_BULK          1       /* Bulk write command */
-#define REDIS_CMD_INLINE        2       /* Inline command */
-/* REDIS_CMD_DENYOOM reserves a longer comment: all the commands marked with
-   this flags will return an error when the 'maxmemory' option is set in the
-   config file and the server is using more than maxmemory bytes of memory.
-   In short this commands are denied on low memory conditions. */
-#define REDIS_CMD_DENYOOM       4
-#define REDIS_CMD_FORCE_REPLICATION 8 /* Force replication even if dirty is 0 */
+/* Command flags:
+ *   REDIS_CMD_DENYOOM:
+ *     Commands marked with this flag will return an error when 'maxmemory' is
+ *     set and the server is using more than 'maxmemory' bytes of memory.
+ *     In short: commands with this flag are denied on low memory conditions.
+ *   REDIS_CMD_FORCE_REPLICATION:
+ *     Force replication even if dirty is 0. */
+#define REDIS_CMD_DENYOOM 4
+#define REDIS_CMD_FORCE_REPLICATION 8

 
 /* Object types */
 #define REDIS_STRING 0
 
 /* Object types */
 #define REDIS_STRING 0


@@ -143,11 +143,17 @@
 #define REDIS_BLOCKED 16    /* The client is waiting in a blocking operation */
 #define REDIS_IO_WAIT 32    /* The client is waiting for Virtual Memory I/O */
 #define REDIS_DIRTY_CAS 64  /* Watched keys modified. EXEC will fail. */
 #define REDIS_BLOCKED 16    /* The client is waiting in a blocking operation */
 #define REDIS_IO_WAIT 32    /* The client is waiting for Virtual Memory I/O */
 #define REDIS_DIRTY_CAS 64  /* Watched keys modified. EXEC will fail. */

+#define REDIS_CLOSE_AFTER_REPLY 128 /* Close after writing entire reply. */
+
+/* Client request types */
+#define REDIS_REQ_INLINE 1
+#define REDIS_REQ_MULTIBULK 2

 
 /* Slave replication state - slave side */
 #define REDIS_REPL_NONE 0   /* No active replication */
 #define REDIS_REPL_CONNECT 1    /* Must connect to master */
 
 /* Slave replication state - slave side */
 #define REDIS_REPL_NONE 0   /* No active replication */
 #define REDIS_REPL_CONNECT 1    /* Must connect to master */

-#define REDIS_REPL_CONNECTED 2  /* Connected to master */
+#define REDIS_REPL_TRANSFER 2    /* Receiving .rdb from master */
+#define REDIS_REPL_CONNECTED 3  /* Connected to master */

 
 /* Slave replication state - from the point of view of master
  * Note that in SEND_BULK and ONLINE state the slave receives new updates
 
 /* Slave replication state - from the point of view of master
  * Note that in SEND_BULK and ONLINE state the slave receives new updates


@@ -197,6 +203,13 @@
 #define REDIS_OP_DIFF 1
 #define REDIS_OP_INTER 2
 
 #define REDIS_OP_DIFF 1
 #define REDIS_OP_INTER 2
 

+/* Redis maxmemory strategies */
+#define REDIS_MAXMEMORY_VOLATILE_LRU 0
+#define REDIS_MAXMEMORY_VOLATILE_TTL 1
+#define REDIS_MAXMEMORY_VOLATILE_RANDOM 2
+#define REDIS_MAXMEMORY_ALLKEYS_LRU 3
+#define REDIS_MAXMEMORY_ALLKEYS_RANDOM 4
+

 /* We can print the stacktrace, so our assert is defined this way: */
 #define redisAssert(_e) ((_e)?(void)0 : (_redisAssert(#_e,__FILE__,__LINE__),_exit(1)))
 #define redisPanic(_e) _redisPanic(#_e,__FILE__,__LINE__),_exit(1)
 /* We can print the stacktrace, so our assert is defined this way: */
 #define redisAssert(_e) ((_e)?(void)0 : (_redisAssert(#_e,__FILE__,__LINE__),_exit(1)))
 #define redisPanic(_e) _redisPanic(#_e,__FILE__,__LINE__),_exit(1)


@@ -210,6 +223,8 @@ void _redisPanic(char *msg, char *file, int line);
 /* A redis object, that is a type able to hold a string / list / set */
 
 /* The actual Redis Object */
 /* A redis object, that is a type able to hold a string / list / set */
 
 /* The actual Redis Object */

+#define REDIS_LRU_CLOCK_MAX ((1<<21)-1) /* Max value of obj->lru */
+#define REDIS_LRU_CLOCK_RESOLUTION 10 /* LRU clock resolution in seconds */

 typedef struct redisObject {
     unsigned type:4;
     unsigned storage:2;     /* REDIS_VM_MEMORY or REDIS_VM_SWAPPING */
 typedef struct redisObject {
     unsigned type:4;
     unsigned storage:2;     /* REDIS_VM_MEMORY or REDIS_VM_SWAPPING */


@@ -284,10 +299,11 @@ typedef struct redisClient {
     redisDb *db;
     int dictid;
     sds querybuf;
     redisDb *db;
     int dictid;
     sds querybuf;

-    robj **argv, **mbargv;
-    int argc, mbargc;
-    long bulklen;            /* bulk read len. -1 if not in bulk read mode */
-    int multibulk;          /* multi bulk command format active */
+    int argc;
+    robj **argv;
+    int reqtype;
+    int multibulklen;       /* number of multi bulk arguments left to read */
+    long bulklen;           /* length of bulk argument in multi bulk request */

     list *reply;
     int sentlen;
     time_t lastinteraction; /* time of the last interaction, used for timeout */
     list *reply;
     int sentlen;
     time_t lastinteraction; /* time of the last interaction, used for timeout */


@@ -309,6 +325,10 @@ typedef struct redisClient {
     list *watched_keys;     /* Keys WATCHED for MULTI/EXEC CAS */
     dict *pubsub_channels;  /* channels a client is interested in (SUBSCRIBE) */
     list *pubsub_patterns;  /* patterns a client is interested in (SUBSCRIBE) */
     list *watched_keys;     /* Keys WATCHED for MULTI/EXEC CAS */
     dict *pubsub_channels;  /* channels a client is interested in (SUBSCRIBE) */
     list *pubsub_patterns;  /* patterns a client is interested in (SUBSCRIBE) */

+
+    /* Response buffer */
+    int bufpos;
+    char buf[REDIS_REPLY_CHUNK_BYTES];

 } redisClient;
 
 struct saveparam {
 } redisClient;
 
 struct saveparam {


@@ -332,22 +352,29 @@ struct sharedObjectsStruct {
 struct redisServer {
     pthread_t mainthread;
     int port;
 struct redisServer {
     pthread_t mainthread;
     int port;

-    int fd;
+    char *bindaddr;
+    char *unixsocket;
+    int ipfd;
+    int sofd;

     redisDb *db;
     long long dirty;            /* changes to DB from the last save */
     long long dirty_before_bgsave; /* used to restore dirty on failed BGSAVE */
     list *clients;
     redisDb *db;
     long long dirty;            /* changes to DB from the last save */
     long long dirty_before_bgsave; /* used to restore dirty on failed BGSAVE */
     list *clients;

+    dict *commands;             /* Command table hahs table */
+    /* Fast pointers to often looked up command */
+    struct redisCommand *delCommand, *multiCommand;

     list *slaves, *monitors;
     char neterr[ANET_ERR_LEN];
     aeEventLoop *el;
     int cronloops;              /* number of times the cron function run */
     list *slaves, *monitors;
     char neterr[ANET_ERR_LEN];
     aeEventLoop *el;
     int cronloops;              /* number of times the cron function run */

-    list *objfreelist;          /* A list of freed objects to avoid malloc() */
-    time_t lastsave;            /* Unix time of last save succeeede */
+    time_t lastsave;                /* Unix time of last save succeeede */

     /* Fields used only for stats */
     /* Fields used only for stats */

-    time_t stat_starttime;         /* server start time */
-    long long stat_numcommands;    /* number of processed commands */
-    long long stat_numconnections; /* number of connections received */
-    long long stat_expiredkeys;   /* number of expired keys */
+    time_t stat_starttime;          /* server start time */
+    long long stat_numcommands;     /* number of processed commands */
+    long long stat_numconnections;  /* number of connections received */
+    long long stat_expiredkeys;     /* number of expired keys */
+    long long stat_keyspace_hits;   /* number of successful lookups of keys */
+    long long stat_keyspace_misses; /* number of failed lookups of keys */

     /* Configuration */
     int verbosity;
     int glueoutputbuf;
     /* Configuration */
     int verbosity;
     int glueoutputbuf;


@@ -369,7 +396,6 @@ struct redisServer {
     struct saveparam *saveparams;
     int saveparamslen;
     char *logfile;
     struct saveparam *saveparams;
     int saveparamslen;
     char *logfile;

-    char *bindaddr;

     char *dbfilename;
     char *appendfilename;
     char *requirepass;
     char *dbfilename;
     char *appendfilename;
     char *requirepass;


@@ -377,13 +403,24 @@ struct redisServer {
     int activerehashing;
     /* Replication related */
     int isslave;
     int activerehashing;
     /* Replication related */
     int isslave;

+    /* Slave specific fields */

     char *masterauth;
     char *masterhost;
     int masterport;
     redisClient *master;    /* client that is master for this slave */
     char *masterauth;
     char *masterhost;
     int masterport;
     redisClient *master;    /* client that is master for this slave */

-    int replstate;
+    int replstate;          /* replication status if the instance is a slave */
+    off_t repl_transfer_left;  /* bytes left reading .rdb  */
+    int repl_transfer_s;    /* slave -> master SYNC socket */
+    int repl_transfer_fd;   /* slave -> master SYNC temp file descriptor */
+    char *repl_transfer_tmpfile; /* slave-> master SYNC temp file name */
+    time_t repl_transfer_lastio; /* unix time of the latest read, for timeout */
+    int repl_serve_stale_data; /* Serve stale data when link is down? */
+    /* Limits */

     unsigned int maxclients;
     unsigned long long maxmemory;
     unsigned int maxclients;
     unsigned long long maxmemory;

+    int maxmemory_policy;
+    int maxmemory_samples;
+    /* Blocked clients */

     unsigned int blpop_blocked_clients;
     unsigned int vm_blocked_clients;
     /* Sort parameters - qsort_r() is only available under BSD so we
     unsigned int blpop_blocked_clients;
     unsigned int vm_blocked_clients;
     /* Sort parameters - qsort_r() is only available under BSD so we


@@ -419,7 +456,6 @@ struct redisServer {
     list *io_processed; /* List of VM I/O jobs already processed */
     list *io_ready_clients; /* Clients ready to be unblocked. All keys loaded */
     pthread_mutex_t io_mutex; /* lock to access io_jobs/io_done/io_thread_job */
     list *io_processed; /* List of VM I/O jobs already processed */
     list *io_ready_clients; /* Clients ready to be unblocked. All keys loaded */
     pthread_mutex_t io_mutex; /* lock to access io_jobs/io_done/io_thread_job */

-    pthread_mutex_t obj_freelist_mutex; /* safe redis objects creation/free */

     pthread_mutex_t io_swapfile_mutex; /* So we can lseek + write */
     pthread_attr_t io_threads_attr; /* attributes for threads creation */
     int io_active_threads; /* Number of running I/O threads */
     pthread_mutex_t io_swapfile_mutex; /* So we can lseek + write */
     pthread_attr_t io_threads_attr; /* attributes for threads creation */
     int io_active_threads; /* Number of running I/O threads */


@@ -485,13 +521,14 @@ typedef struct _redisSortOperation {
 } redisSortOperation;
 
 /* ZSETs use a specialized version of Skiplists */
 } redisSortOperation;
 
 /* ZSETs use a specialized version of Skiplists */

-

 typedef struct zskiplistNode {
 typedef struct zskiplistNode {

-    struct zskiplistNode **forward;
-    struct zskiplistNode *backward;
-    unsigned int *span;
-    double score;

     robj *obj;
     robj *obj;

+    double score;
+    struct zskiplistNode *backward;
+    struct zskiplistLevel {
+        struct zskiplistNode *forward;
+        unsigned int span;
+    } level[];

 } zskiplistNode;
 
 typedef struct zskiplist {
 } zskiplistNode;
 
 typedef struct zskiplist {


@@ -588,20 +625,35 @@ void resetClient(redisClient *c);
 void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask);
 void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReply(redisClient *c, robj *obj);
 void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask);
 void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReply(redisClient *c, robj *obj);

+void *addDeferredMultiBulkLength(redisClient *c);
+void setDeferredMultiBulkLength(redisClient *c, void *node, long length);

 void addReplySds(redisClient *c, sds s);
 void processInputBuffer(redisClient *c);
 void addReplySds(redisClient *c, sds s);
 void processInputBuffer(redisClient *c);

-void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask);
+void acceptTcpHandler(aeEventLoop *el, int fd, void *privdata, int mask);
+void acceptUnixHandler(aeEventLoop *el, int fd, void *privdata, int mask);

 void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReplyBulk(redisClient *c, robj *obj);
 void addReplyBulkCString(redisClient *c, char *s);
 void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReply(redisClient *c, robj *obj);
 void addReplySds(redisClient *c, sds s);
 void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReplyBulk(redisClient *c, robj *obj);
 void addReplyBulkCString(redisClient *c, char *s);
 void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReply(redisClient *c, robj *obj);
 void addReplySds(redisClient *c, sds s);

+void addReplyError(redisClient *c, char *err);
+void addReplyStatus(redisClient *c, char *status);

 void addReplyDouble(redisClient *c, double d);
 void addReplyLongLong(redisClient *c, long long ll);
 void addReplyDouble(redisClient *c, double d);
 void addReplyLongLong(redisClient *c, long long ll);

-void addReplyUlong(redisClient *c, unsigned long ul);
+void addReplyMultiBulkLen(redisClient *c, long length);

 void *dupClientReplyValue(void *o);
 
 void *dupClientReplyValue(void *o);
 

+#ifdef __GNUC__
+void addReplyErrorFormat(redisClient *c, const char *fmt, ...)
+    __attribute__((format(printf, 2, 3)));
+void addReplyStatusFormat(redisClient *c, const char *fmt, ...)
+    __attribute__((format(printf, 2, 3)));
+#else
+void addReplyErrorFormat(redisClient *c, const char *fmt, ...);
+void addReplyStatusFormat(redisClient *c, const char *fmt, ...);
+#endif
+

 /* List data type */
 void listTypeTryConversion(robj *subject, robj *value);
 void listTypePush(robj *subject, robj *value, int where);
 /* List data type */
 void listTypeTryConversion(robj *subject, robj *value);
 void listTypePush(robj *subject, robj *value, int where);


@@ -641,7 +693,6 @@ robj *dupStringObject(robj *o);
 robj *tryObjectEncoding(robj *o);
 robj *getDecodedObject(robj *o);
 size_t stringObjectLen(robj *o);
 robj *tryObjectEncoding(robj *o);
 robj *getDecodedObject(robj *o);
 size_t stringObjectLen(robj *o);

-int tryFreeOneObjectFromFreelist(void);

 robj *createStringObjectFromLongLong(long long value);
 robj *createListObject(void);
 robj *createZiplistObject(void);
 robj *createStringObjectFromLongLong(long long value);
 robj *createListObject(void);
 robj *createZiplistObject(void);


@@ -657,12 +708,23 @@ int getLongLongFromObject(robj *o, long long *target);
 char *strEncoding(int encoding);
 int compareStringObjects(robj *a, robj *b);
 int equalStringObjects(robj *a, robj *b);
 char *strEncoding(int encoding);
 int compareStringObjects(robj *a, robj *b);
 int equalStringObjects(robj *a, robj *b);

+unsigned long estimateObjectIdleTime(robj *o);
+
+/* Synchronous I/O with timeout */
+int syncWrite(int fd, char *ptr, ssize_t size, int timeout);
+int syncRead(int fd, char *ptr, ssize_t size, int timeout);
+int syncReadLine(int fd, char *ptr, ssize_t size, int timeout);
+int fwriteBulkString(FILE *fp, char *s, unsigned long len);
+int fwriteBulkDouble(FILE *fp, double d);
+int fwriteBulkLongLong(FILE *fp, long long l);
+int fwriteBulkObject(FILE *fp, robj *obj);

 
 /* Replication */
 void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc);
 void replicationFeedMonitors(list *monitors, int dictid, robj **argv, int argc);
 int syncWithMaster(void);
 void updateSlavesWaitingBgsave(int bgsaveerr);
 
 /* Replication */
 void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc);
 void replicationFeedMonitors(list *monitors, int dictid, robj **argv, int argc);
 int syncWithMaster(void);
 void updateSlavesWaitingBgsave(int bgsaveerr);

+void replicationCron(void);

 
 /* RDB persistence */
 int rdbLoad(char *filename);
 
 /* RDB persistence */
 int rdbLoad(char *filename);


@@ -688,13 +750,14 @@ void backgroundRewriteDoneHandler(int statloc);
 /* Sorted sets data type */
 zskiplist *zslCreate(void);
 void zslFree(zskiplist *zsl);
 /* Sorted sets data type */
 zskiplist *zslCreate(void);
 void zslFree(zskiplist *zsl);

-void zslInsert(zskiplist *zsl, double score, robj *obj);
+zskiplistNode *zslInsert(zskiplist *zsl, double score, robj *obj);

 
 /* Core functions */
 void freeMemoryIfNeeded(void);
 int processCommand(redisClient *c);
 void setupSigSegvAction(void);
 
 /* Core functions */
 void freeMemoryIfNeeded(void);
 int processCommand(redisClient *c);
 void setupSigSegvAction(void);

-struct redisCommand *lookupCommand(char *name);
+struct redisCommand *lookupCommand(sds name);
+struct redisCommand *lookupCommandByCString(char *s);

 void call(redisClient *c, struct redisCommand *cmd);
 int prepareForShutdown();
 void redisLog(int level, const char *fmt, ...);
 void call(redisClient *c, struct redisCommand *cmd);
 int prepareForShutdown();
 void redisLog(int level, const char *fmt, ...);


@@ -702,6 +765,7 @@ void usage();
 void updateDictResizePolicy(void);
 int htNeedsResize(dict *dict);
 void oom(const char *msg);
 void updateDictResizePolicy(void);
 int htNeedsResize(dict *dict);
 void oom(const char *msg);

+void populateCommandTable(void);

 
 /* Virtual Memory */
 void vmInit(void);
 
 /* Virtual Memory */
 void vmInit(void);


@@ -875,6 +939,7 @@ void zaddCommand(redisClient *c);
 void zincrbyCommand(redisClient *c);
 void zrangeCommand(redisClient *c);
 void zrangebyscoreCommand(redisClient *c);
 void zincrbyCommand(redisClient *c);
 void zrangeCommand(redisClient *c);
 void zrangebyscoreCommand(redisClient *c);

+void zrevrangebyscoreCommand(redisClient *c);

 void zcountCommand(redisClient *c);
 void zrevrangeCommand(redisClient *c);
 void zcardCommand(redisClient *c);
 void zcountCommand(redisClient *c);
 void zrevrangeCommand(redisClient *c);
 void zcardCommand(redisClient *c);