add machamp script

Check the validity of the value before performing the create operation,
prevents new data from being generated even if the request fails to execute.

Co-authored-by: Oran Agra <oran@redislabs.com>
Co-authored-by: chendianqiang <chendianqiang@meituan.com>
Co-authored-by: Binbin <binloveplay1314@qq.com>

build.yaml

@@ -0,0 +1,53 @@
+# Doc: https://wiki.sc-corp.net/pages/viewpage.action?pageId=121500284
+version: 1
+machamp:
+  keydb-build:
+    # Optional - build counter is linked to the build def
+    tag_template: 0.0.%build.counter%
+    # Optional - value in seconds before a build is terminated, default is 3600 seconds
+    timeout: 3600
+    # Optional - update ghe or not, default to true
+    update_ghe: true
+    code_coverage: false
+    # Required
+    steps:
+      make-build:
+        type: cmd
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./machamp_scripts/build.sh
+      tls-test:
+        type: cmd
+        parent: make-build
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./runtest --clients 4 --verbose 
+      cluster-test:
+        type: cmd
+        parent: make-build
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./runtest-cluster 
+      sentinel-test:
+        type: cmd
+        parent: make-build
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./runtest-sentinel
+      module-test:
+        type: cmd
+        parent: make-build
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./runtest-moduleapi
+      rotation-test:
+        type: cmd
+        parent: make-build
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./runtest-rotation
+      keyproxy-test:
+        type: cmd
+        parent: make-build
+        command: ./runtest-keyproxy --tls
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c

ci.yaml

@@ -0,0 +1,16 @@
+# Doc: https://wiki.sc-corp.net/display/TOOL/ci.yaml+User+Guide
+version: 1
+on:
+  pull_request:
+    - workflows:
+      # All builds that use machamp should use the defined `backend_workflow`
+      - workflow_type: backend_workflow
+        # references a build defined in build.yaml
+        build_name: keydb-build
+        arch_types: ["amd64", "arm64"]
+  push:
+    - branches: [main]
+      workflows:
+        - workflow_type: backend_workflow
+          build_name: keydb-build
+          arch_types: ["amd64", "arm64"]

machamp_scripts/build.sh

@@ -0,0 +1,8 @@
+#!/bin/bash
+
+# make the build
+git submodule init && git submodule update
+make BUILD_TLS=yes -j$(nproc) KEYDB_CFLAGS='-Werror' KEYDB_CXXFLAGS='-Werror'
+
+# gen-cert
+./utils/gen-test-certs.sh

src/IStorage.h

@@ -31,6 +31,7 @@ public:
     virtual void retrieve(const char *key, size_t cchKey, callbackSingle fn) const = 0;
     virtual size_t clear() = 0;
     virtual bool enumerate(callback fn) const = 0;
+    virtual bool enumerate_hashslot(callback fn, unsigned int hashslot) const = 0;
     virtual size_t count() const = 0;
 
     virtual void bulkInsert(char **rgkeys, size_t *rgcbkeys, char **rgvals, size_t *rgcbvals, size_t celem) {

src/Makefile

@@ -145,7 +145,9 @@ FINAL_LDFLAGS=$(LDFLAGS) $(KEYDB_LDFLAGS) $(DEBUG)
 FINAL_LIBS+=-lm -lz -lcrypto -lbz2 -lzstd -llz4 -lsnappy
 
 ifneq ($(uname_S),Darwin)
+    ifneq ($(uname_S),FreeBSD)
 	FINAL_LIBS+=-latomic
+    endif
 endif
 # Linux ARM32 needs -latomic at linking time
 ifneq (,$(findstring armv,$(uname_M)))
@@ -211,7 +213,13 @@ ifeq ($(uname_S),NetBSD)
 else
 ifeq ($(uname_S),FreeBSD)
 	# FreeBSD
-	FINAL_LIBS+= -lpthread -lexecinfo
+	FINAL_LIBS+= -lpthread -luuid -lexecinfo
+	FINAL_CFLAGS+= -I/usr/local/include
+	FINAL_CXXFLAGS+= -I/usr/local/include
+	FINAL_LDFLAGS+= -L/usr/local/lib
+	ifeq ($(USE_BACKTRACE),yes)
+	    FINAL_CFLAGS+= -DUSE_BACKTRACE
+	endif
 else
 ifeq ($(uname_S),DragonFly)
 	# DragonFly

src/StorageCache.cpp

@@ -21,7 +21,7 @@ dictType dbStorageCacheType = {
 StorageCache::StorageCache(IStorage *storage, bool fCache)
         : m_spstorage(storage)
 {
-    if (fCache)
+    if (!g_pserver->flash_disable_key_cache && fCache)
         m_pdict = dictCreate(&dbStorageCacheType, nullptr);
 }
 

src/StorageCache.h

@@ -49,6 +49,7 @@ public:
     void expand(uint64_t slots);
 
     bool enumerate(IStorage::callback fn) const { return m_spstorage->enumerate(fn); }
+    bool enumerate_hashslot(IStorage::callback fn, unsigned int hashslot) const { return m_spstorage->enumerate_hashslot(fn, hashslot); }
 
     void beginWriteBatch();
     void endWriteBatch() { m_spstorage->endWriteBatch(); }

src/config.cpp

@@ -2806,7 +2806,7 @@ standardConfig configs[] = {
     createBoolConfig("appendonly", NULL, MODIFIABLE_CONFIG, g_pserver->aof_enabled, 0, NULL, updateAppendonly),
     createBoolConfig("cluster-allow-reads-when-down", NULL, MODIFIABLE_CONFIG, g_pserver->cluster_allow_reads_when_down, 0, NULL, NULL),
     createBoolConfig("delete-on-evict", NULL, MODIFIABLE_CONFIG, cserver.delete_on_evict, 0, NULL, NULL),
-    createBoolConfig("use-fork", NULL, IMMUTABLE_CONFIG, cserver.fForkBgSave, 0, NULL, NULL),
+    createBoolConfig("use-fork", NULL, IMMUTABLE_CONFIG, cserver.fForkBgSave, 1, NULL, NULL),
     createBoolConfig("io-threads-do-reads", NULL, IMMUTABLE_CONFIG, fDummy, 0, NULL, NULL),
     createBoolConfig("time-thread-priority", NULL, IMMUTABLE_CONFIG, cserver.time_thread_priority, 0, NULL, NULL),
     createBoolConfig("prefetch-enabled", NULL, MODIFIABLE_CONFIG, g_pserver->prefetch_enabled, 1, NULL, NULL),
@@ -2912,6 +2912,7 @@ standardConfig configs[] = {
     createLongLongConfig("repl-backlog-size", NULL, MODIFIABLE_CONFIG, 1, LLONG_MAX, g_pserver->repl_backlog_size, 1024*1024, MEMORY_CONFIG, NULL, updateReplBacklogSize), /* Default: 1mb */
     createLongLongConfig("repl-backlog-disk-reserve", NULL, IMMUTABLE_CONFIG, 0, LLONG_MAX, cserver.repl_backlog_disk_size, 0, MEMORY_CONFIG, NULL, NULL),
     createLongLongConfig("max-snapshot-slip", NULL, MODIFIABLE_CONFIG, 0, 5000, g_pserver->snapshot_slip, 400, 0, NULL, NULL),
+    createLongLongConfig("max-rand-count", NULL, MODIFIABLE_CONFIG, 0, LONG_MAX/2, g_pserver->rand_total_threshold, LONG_MAX/2, 0, NULL, NULL),
 
     /* Unsigned Long Long configs */
     createULongLongConfig("maxmemory", NULL, MODIFIABLE_CONFIG, 0, LLONG_MAX, g_pserver->maxmemory, 0, MEMORY_CONFIG, NULL, updateMaxmemory),
@@ -2939,6 +2940,7 @@ standardConfig configs[] = {
     createBoolConfig("allow-write-during-load", NULL, MODIFIABLE_CONFIG, g_pserver->fWriteDuringActiveLoad, 0, NULL, NULL),
     createBoolConfig("force-backlog-disk-reserve", NULL, MODIFIABLE_CONFIG, cserver.force_backlog_disk, 0, NULL, NULL),
     createBoolConfig("soft-shutdown", NULL, MODIFIABLE_CONFIG, g_pserver->config_soft_shutdown, 0, NULL, NULL),
+    createBoolConfig("flash-disable-key-cache", NULL, MODIFIABLE_CONFIG, g_pserver->flash_disable_key_cache, 0, NULL, NULL),
     createSizeTConfig("semi-ordered-set-bucket-size", NULL, MODIFIABLE_CONFIG, 0, 1024, g_semiOrderedSetTargetBucketSize, 0, INTEGER_CONFIG, NULL, NULL),
 
 #ifdef USE_OPENSSL

src/db.cpp

@@ -1737,6 +1737,9 @@ int dbSwapDatabases(int id1, int id2) {
         id2 < 0 || id2 >= cserver.dbnum) return C_ERR;
     if (id1 == id2) return C_OK;
     std::swap(g_pserver->db[id1], g_pserver->db[id2]);
+    
+    //swap db's id too, otherwise db does not match its id
+    std::swap(g_pserver->db[id1]->id, g_pserver->db[id2]->id);
 
     /* Note that we don't swap blocking_keys,
      * ready_keys and watched_keys, since we want clients to
@@ -1766,7 +1769,7 @@ int dbSwapDatabases(int id1, int id2) {
 
 /* SWAPDB db1 db2 */
 void swapdbCommand(client *c) {
-    int id1, id2;
+    int id1, id2, oriId;
 
     /* Not allowed in cluster mode: we have just DB 0 there. */
     if (g_pserver->cluster_enabled) {
@@ -1783,6 +1786,9 @@ void swapdbCommand(client *c) {
         "invalid second DB index") != C_OK)
         return;
 
+    //get client's original db's id
+    oriId=c->db->id;
+
     /* Swap... */
     if (dbSwapDatabases(id1,id2) == C_ERR) {
         addReplyError(c,"DB index is out of range");
@@ -1791,6 +1797,10 @@ void swapdbCommand(client *c) {
         RedisModuleSwapDbInfo si = {REDISMODULE_SWAPDBINFO_VERSION,(int32_t)id1,(int32_t)id2};
         moduleFireServerEvent(REDISMODULE_EVENT_SWAPDB,0,&si);
         g_pserver->dirty++;
+
+	//set client's db to original db
+        c->db=g_pserver->db[oriId];
+
         addReply(c,shared.ok);
     }
 }
@@ -1931,6 +1941,7 @@ void setExpire(client *c, redisDb *db, robj *key, expireEntry &&e)
  * is associated with this key (i.e. the key is non volatile) */
 expireEntry *redisDbPersistentDataSnapshot::getExpire(const char *key) {
     /* No expire? return ASAP */
+    std::unique_lock<fastlock> ul(g_expireLock);
     if (expireSize() == 0)
         return nullptr;
 
@@ -2479,25 +2490,28 @@ void slotToKeyUpdateKeyCore(const char *key, size_t keylen, int add) {
     serverAssert(GlobalLocksAcquired());
 
     unsigned int hashslot = keyHashSlot(key,keylen);
-    unsigned char buf[64];
-    unsigned char *indexed = buf;
-
     g_pserver->cluster->slots_keys_count[hashslot] += add ? 1 : -1;
-    if (keylen+2 > 64) indexed = (unsigned char*)zmalloc(keylen+2, MALLOC_SHARED);
-    indexed[0] = (hashslot >> 8) & 0xff;
-    indexed[1] = hashslot & 0xff;
-    memcpy(indexed+2,key,keylen);
-    int fModified = false;
-    if (add) {
-        fModified = raxInsert(g_pserver->cluster->slots_to_keys,indexed,keylen+2,NULL,NULL);
-    } else {
-        fModified = raxRemove(g_pserver->cluster->slots_to_keys,indexed,keylen+2,NULL);
+
+    if (g_pserver->m_pstorageFactory == nullptr) {
+        unsigned char buf[64];
+        unsigned char *indexed = buf;
+
+        if (keylen+2 > 64) indexed = (unsigned char*)zmalloc(keylen+2, MALLOC_SHARED);
+        indexed[0] = (hashslot >> 8) & 0xff;
+        indexed[1] = hashslot & 0xff;
+        memcpy(indexed+2,key,keylen);
+        int fModified = false;
+        if (add) {
+            fModified = raxInsert(g_pserver->cluster->slots_to_keys,indexed,keylen+2,NULL,NULL);
+        } else {
+            fModified = raxRemove(g_pserver->cluster->slots_to_keys,indexed,keylen+2,NULL);
+        }
+        // This assert is disabled when a snapshot depth is >0 because prepOverwriteForSnapshot will add in a tombstone,
+        //  this prevents ensure from adding the key to the dictionary which means the caller isn't aware we're already tracking
+        //  the key.
+        serverAssert(fModified || g_pserver->db[0]->snapshot_depth() > 0);
+        if (indexed != buf) zfree(indexed);
     }
-    // This assert is disabled when a snapshot depth is >0 because prepOverwriteForSnapshot will add in a tombstone,
-    //  this prevents ensure from adding the key to the dictionary which means the caller isn't aware we're already tracking
-    //  the key.
-    serverAssert(fModified || g_pserver->db[0]->snapshot_depth() > 0);
-    if (indexed != buf) zfree(indexed);
 }
 
 void slotToKeyAdd(sds key) {
@@ -2533,47 +2547,67 @@ void slotToKeyFlush(int async) {
  * New objects are returned to represent keys, it's up to the caller to
  * decrement the reference count to release the keys names. */
 unsigned int getKeysInSlot(unsigned int hashslot, robj **keys, unsigned int count) {
-    raxIterator iter;
-    int j = 0;
-    unsigned char indexed[2];
+    if (g_pserver->m_pstorageFactory != nullptr) {
+        int j = 0;
+        g_pserver->db[0]->getStorageCache()->enumerate_hashslot([&](const char *key, size_t cchKey, const void *, size_t )->bool {
+            keys[j++] = createStringObject(key, cchKey);
+            return --count;
+        }, hashslot);
+        return j;
+    } else {
+        raxIterator iter;
+        int j = 0;
+        unsigned char indexed[2];
 
-    indexed[0] = (hashslot >> 8) & 0xff;
-    indexed[1] = hashslot & 0xff;
-    raxStart(&iter,g_pserver->cluster->slots_to_keys);
-    raxSeek(&iter,">=",indexed,2);
-    while(count-- && raxNext(&iter)) {
-        if (iter.key[0] != indexed[0] || iter.key[1] != indexed[1]) break;
-        keys[j++] = createStringObject((char*)iter.key+2,iter.key_len-2);
+        indexed[0] = (hashslot >> 8) & 0xff;
+        indexed[1] = hashslot & 0xff;
+        raxStart(&iter,g_pserver->cluster->slots_to_keys);
+        raxSeek(&iter,">=",indexed,2);
+        while(count-- && raxNext(&iter)) {
+            if (iter.key[0] != indexed[0] || iter.key[1] != indexed[1]) break;
+            keys[j++] = createStringObject((char*)iter.key+2,iter.key_len-2);
+        }
+        raxStop(&iter);
+        return j;
     }
-    raxStop(&iter);
-    return j;
 }
 
 /* Remove all the keys in the specified hash slot.
  * The number of removed items is returned. */
 unsigned int delKeysInSlot(unsigned int hashslot) {
     serverAssert(GlobalLocksAcquired());
-    
-    raxIterator iter;
-    int j = 0;
-    unsigned char indexed[2];
+    if (g_pserver->m_pstorageFactory != nullptr) {
+        int j = 0;
+        g_pserver->db[0]->getStorageCache()->enumerate_hashslot([&](const char *key, size_t cchKey, const void *, size_t )->bool {
+            robj *keyobj = createStringObject(key, cchKey);
+            dbDelete(g_pserver->db[0], keyobj);
+            decrRefCount(keyobj);
+            j++;
+            return true;
+        }, hashslot);
+        return j;
+    } else {
+        raxIterator iter;
+        int j = 0;
+        unsigned char indexed[2];
 
-    indexed[0] = (hashslot >> 8) & 0xff;
-    indexed[1] = hashslot & 0xff;
-    raxStart(&iter,g_pserver->cluster->slots_to_keys);
-    while(g_pserver->cluster->slots_keys_count[hashslot]) {
-        raxSeek(&iter,">=",indexed,2);
-        raxNext(&iter);
+        indexed[0] = (hashslot >> 8) & 0xff;
+        indexed[1] = hashslot & 0xff;
+        raxStart(&iter,g_pserver->cluster->slots_to_keys);
+        while(g_pserver->cluster->slots_keys_count[hashslot]) {
+            raxSeek(&iter,">=",indexed,2);
+            raxNext(&iter);
 
-        auto count = g_pserver->cluster->slots_keys_count[hashslot];
-        robj *key = createStringObject((char*)iter.key+2,iter.key_len-2);
-        dbDelete(g_pserver->db[0],key);
-        serverAssert(count > g_pserver->cluster->slots_keys_count[hashslot]);   // we should have deleted something or we will be in an infinite loop
-        decrRefCount(key);
-        j++;
+            auto count = g_pserver->cluster->slots_keys_count[hashslot];
+            robj *key = createStringObject((char*)iter.key+2,iter.key_len-2);
+            dbDelete(g_pserver->db[0],key);
+            serverAssert(count > g_pserver->cluster->slots_keys_count[hashslot]);   // we should have deleted something or we will be in an infinite loop
+            decrRefCount(key);
+            j++;
+        }
+        raxStop(&iter);
+        return j;
     }
-    raxStop(&iter);
-    return j;
 }
 
 unsigned int countKeysInSlot(unsigned int hashslot) {
@@ -2727,8 +2761,11 @@ void redisDbPersistentData::clear(void(callback)(void*))
         m_cnewKeysPending = 0;
         m_fAllChanged++;
     }
+    {
+    std::unique_lock<fastlock> ul(g_expireLock);
     delete m_setexpire;
     m_setexpire = new (MALLOC_LOCAL) expireset();
+    }
     if (m_spstorage != nullptr)
         m_spstorage->clear(callback);
     dictEmpty(m_pdictTombstone,callback);
@@ -2861,7 +2898,8 @@ LNotFound:
             {
                 dictAdd(m_pdict, sdsNewKey, o);
                 o->SetFExpires(spexpire != nullptr);
-
+                
+                std::unique_lock<fastlock> ul(g_expireLock);
                 if (spexpire != nullptr)
                 {
                     auto itr = m_setexpire->find(sdsKey);
@@ -2884,6 +2922,7 @@ LNotFound:
     if (*pde != nullptr && dictGetVal(*pde) != nullptr)
     {
         robj *o = (robj*)dictGetVal(*pde);
+        std::unique_lock<fastlock> ul(g_expireLock);
         serverAssert(o->FExpires() == (m_setexpire->find(sdsKey) != m_setexpire->end()));
     }
 }
@@ -3256,6 +3295,7 @@ std::unique_ptr<expireEntry> deserializeExpire(sds key, const char *str, size_t
 
 sds serializeStoredObjectAndExpire(redisDbPersistentData *db, const char *key, robj_roptr o)
 {
+    std::unique_lock<fastlock> ul(g_expireLock);
     auto itrExpire = db->setexpire()->find(key);
     const expireEntry *pexpire = nullptr;
     if (itrExpire != db->setexpire()->end())
@@ -3380,6 +3420,7 @@ void redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command
                     dictAdd(m_pdict, sharedKey, o);
                     o->SetFExpires(spexpire != nullptr);
 
+                    std::unique_lock<fastlock> ul(g_expireLock);
                     if (spexpire != nullptr)
                     {
                         auto itr = m_setexpire->find(sharedKey);

src/evict.cpp

@@ -380,7 +380,7 @@ size_t freeMemoryGetNotCountedMemory(void) {
 
     /* also don't count the replication backlog memory
      * that's where the replication clients get their memory from */
-    overhead += g_pserver->repl_backlog_size;
+    overhead += g_pserver->repl_backlog_size - g_pserver->repl_backlog_config_size;
 
     if (g_pserver->aof_state != AOF_OFF) {
         overhead += sdsalloc(g_pserver->aof_buf)+aofRewriteBufferSize();
@@ -771,6 +771,7 @@ int performEvictions(bool fPreSnapshot) {
                 if (db->removeCachedValue(bestkey, &deT)) {
                     mem_freed += splazy->addEntry(db->dictUnsafeKeyOnly(), deT);
                     ckeysFailed = 0;
+		    g_pserver->stat_evictedkeys++;
                 }
                 else {
                     delta = 0;

src/expire.cpp

@@ -481,6 +481,7 @@ void expireSlaveKeys(void) {
     if (slaveKeysWithExpire == NULL ||
         dictSize(slaveKeysWithExpire) == 0) return;
 
+    std::unique_lock<fastlock> ul(g_expireLock);
     int cycles = 0, noexpire = 0;
     mstime_t start = mstime();
     while(1) {

src/object.cpp

@@ -1668,6 +1668,7 @@ robj *deserializeStoredObjectCore(const void *data, size_t cb)
 robj *deserializeStoredObject(const redisDbPersistentData *db, const char *key, const void *data, size_t cb)
 {
     robj *o = deserializeStoredObjectCore(data, cb);
+    std::unique_lock<fastlock> ul(g_expireLock);
     o->SetFExpires(db->setexpire()->exists(key));
     return o;
 }

src/rdb.cpp

@@ -1334,7 +1334,7 @@ int rdbSaveRio(rio *rdb, const redisDbPersistentDataSnapshot **rgpdb, int *error
     if (rdbSaveModulesAux(rdb, REDISMODULE_AUX_BEFORE_RDB) == -1) goto werr;
 
     for (j = 0; j < cserver.dbnum; j++) {
-        const redisDbPersistentDataSnapshot *db = rgpdb[j];
+        const redisDbPersistentDataSnapshot *db = rgpdb != nullptr ? rgpdb[j] : g_pserver->db[j];
         if (db->size() == 0) continue;
 
         /* Write the SELECT DB opcode */
@@ -1715,7 +1715,7 @@ void getTempFileName(char tmpfile[], int tmpfileNum) {
     char tmpfileNumString[214];
 
     /* Generate temp rdb file name using aync-signal safe functions. */
-    int pid_len = ll2string(pid, sizeof(pid), getpid());
+    int pid_len = ll2string(pid, sizeof(pid), g_pserver->in_fork_child ? getppid() : getpid());
     int tmpfileNum_len = ll2string(tmpfileNumString, sizeof(tmpfileNumString), tmpfileNum);
     strcpy(tmpfile, "temp-");
     strncpy(tmpfile+5, pid, pid_len);
@@ -3832,57 +3832,130 @@ int rdbSaveToSlavesSockets(rdbSaveInfo *rsi) {
     }
 
     /* Create the child process. */
-    openChildInfoPipe();
+    if (cserver.fForkBgSave) {
+        pid_t childpid;
+        if ((childpid = redisFork(CHILD_TYPE_RDB)) == 0) {
+            /* Child */
+            int retval, dummy;
+            rio rdb;
 
-    for (int idb = 0; idb < cserver.dbnum; ++idb)
-        args->rgpdb[idb] = g_pserver->db[idb]->createSnapshot(getMvccTstamp(), false /*fOptional*/);
+            rioInitWithFd(&rdb,args->rdb_pipe_write);
 
-    g_pserver->rdbThreadVars.tmpfileNum++;
-    g_pserver->rdbThreadVars.fRdbThreadCancel = false;
-    pthread_attr_t tattr;
-    pthread_attr_init(&tattr);
-    pthread_attr_setstacksize(&tattr, 1 << 23); // 8 MB
-    if (pthread_create(&child, &tattr, rdbSaveToSlavesSocketsThread, args)) {
-        pthread_attr_destroy(&tattr);
-        serverLog(LL_WARNING,"Can't save in background: fork: %s",
-            strerror(errno));
+            redisSetProcTitle("keydb-rdb-to-slaves");
+            redisSetCpuAffinity(g_pserver->bgsave_cpulist);
 
-        /* Undo the state change. The caller will perform cleanup on
-            * all the slaves in BGSAVE_START state, but an early call to
-            * replicationSetupSlaveForFullResync() turned it into BGSAVE_END */
-        listRewind(g_pserver->slaves,&li);
-        while((ln = listNext(&li))) {
-            client *replica = (client*)ln->value;
-            if (replica->replstate == SLAVE_STATE_WAIT_BGSAVE_END) {
-                replica->replstate = SLAVE_STATE_WAIT_BGSAVE_START;
+            retval = rdbSaveRioWithEOFMark(&rdb,nullptr,nullptr,rsi);
+            if (retval == C_OK && rioFlush(&rdb) == 0)
+                retval = C_ERR;
+
+            if (retval == C_OK) {
+                sendChildCowInfo(CHILD_INFO_TYPE_RDB_COW_SIZE, "RDB");
             }
-        }
-        close(args->rdb_pipe_write);
-        close(g_pserver->rdb_pipe_read);
-        zfree(g_pserver->rdb_pipe_conns);
-        close(args->safe_to_exit_pipe);
-        g_pserver->rdb_pipe_conns = NULL;
-        g_pserver->rdb_pipe_numconns = 0;
-        g_pserver->rdb_pipe_numconns_writing = 0;
-        args->rsi.~rdbSaveInfo();
-        zfree(args);
-        closeChildInfoPipe();
-        return C_ERR;
-    }
-    pthread_attr_destroy(&tattr);
-    g_pserver->child_type = CHILD_TYPE_RDB;
 
-    serverLog(LL_NOTICE,"Background RDB transfer started");
-    g_pserver->rdb_save_time_start = time(NULL);
-    serverAssert(!g_pserver->rdbThreadVars.fRdbThreadActive);
-    g_pserver->rdbThreadVars.rdb_child_thread = child;
-    g_pserver->rdbThreadVars.fRdbThreadActive = true;
-    g_pserver->rdb_child_type = RDB_CHILD_TYPE_SOCKET;
-    aePostFunction(g_pserver->rgthreadvar[IDX_EVENT_LOOP_MAIN].el, []{
-        if (aeCreateFileEvent(serverTL->el, g_pserver->rdb_pipe_read, AE_READABLE, rdbPipeReadHandler,NULL) == AE_ERR) {
-            serverPanic("Unrecoverable error creating server.rdb_pipe_read file event.");
+            rioFreeFd(&rdb);
+            /* wake up the reader, tell it we're done. */
+            close(args->rdb_pipe_write);
+            close(g_pserver->rdb_child_exit_pipe); /* close write end so that we can detect the close on the parent. */
+            /* hold exit until the parent tells us it's safe. we're not expecting
+            * to read anything, just get the error when the pipe is closed. */
+            dummy = read(args->safe_to_exit_pipe, pipefds, 1);
+            UNUSED(dummy);
+            exitFromChild((retval == C_OK) ? 0 : 1);
+        } else {
+            /* Parent */
+            close(args->safe_to_exit_pipe);
+            if (childpid == -1) {
+                serverLog(LL_WARNING,"Can't save in background: fork: %s",
+                    strerror(errno));
+
+                /* Undo the state change. The caller will perform cleanup on
+                * all the slaves in BGSAVE_START state, but an early call to
+                * replicationSetupSlaveForFullResync() turned it into BGSAVE_END */
+                listRewind(g_pserver->slaves,&li);
+                while((ln = listNext(&li))) {
+                    client *replica = (client*)ln->value;
+                    if (replica->replstate == SLAVE_STATE_WAIT_BGSAVE_END) {
+                        replica->replstate = SLAVE_STATE_WAIT_BGSAVE_START;
+                    }
+                }
+                close(args->rdb_pipe_write);
+                close(g_pserver->rdb_pipe_read);
+                zfree(g_pserver->rdb_pipe_conns);
+                g_pserver->rdb_pipe_conns = NULL;
+                g_pserver->rdb_pipe_numconns = 0;
+                g_pserver->rdb_pipe_numconns_writing = 0;
+                args->rsi.~rdbSaveInfo();
+                zfree(args);
+            } else {
+                serverLog(LL_NOTICE,"Background RDB transfer started by pid %ld",
+                    (long)childpid);
+                g_pserver->rdb_save_time_start = time(NULL);
+                g_pserver->rdb_child_type = RDB_CHILD_TYPE_SOCKET;
+                g_pserver->rdbThreadVars.fRdbThreadActive = true;
+                updateDictResizePolicy();
+                close(args->rdb_pipe_write); /* close write in parent so that it can detect the close on the child. */
+                aePostFunction(g_pserver->rgthreadvar[IDX_EVENT_LOOP_MAIN].el, []{
+                    if (aeCreateFileEvent(serverTL->el, g_pserver->rdb_pipe_read, AE_READABLE, rdbPipeReadHandler, nullptr) == AE_ERR) {
+                        serverPanic("Unrecoverable error creating g_pserver->rdb_pipe_read file event.");
+                    }
+                });
+            }
+            return (childpid == -1) ? C_ERR : C_OK;
         }
-    });
+    }
+    else {
+        openChildInfoPipe();
+
+        for (int idb = 0; idb < cserver.dbnum; ++idb)
+            args->rgpdb[idb] = g_pserver->db[idb]->createSnapshot(getMvccTstamp(), false /*fOptional*/);
+
+        g_pserver->rdbThreadVars.tmpfileNum++;
+        g_pserver->rdbThreadVars.fRdbThreadCancel = false;
+        pthread_attr_t tattr;
+        pthread_attr_init(&tattr);
+        pthread_attr_setstacksize(&tattr, 1 << 23); // 8 MB
+        if (pthread_create(&child, &tattr, rdbSaveToSlavesSocketsThread, args)) {
+            pthread_attr_destroy(&tattr);
+            serverLog(LL_WARNING,"Can't save in background: fork: %s",
+                strerror(errno));
+
+            /* Undo the state change. The caller will perform cleanup on
+                * all the slaves in BGSAVE_START state, but an early call to
+                * replicationSetupSlaveForFullResync() turned it into BGSAVE_END */
+            listRewind(g_pserver->slaves,&li);
+            while((ln = listNext(&li))) {
+                client *replica = (client*)ln->value;
+                if (replica->replstate == SLAVE_STATE_WAIT_BGSAVE_END) {
+                    replica->replstate = SLAVE_STATE_WAIT_BGSAVE_START;
+                }
+            }
+            close(args->rdb_pipe_write);
+            close(g_pserver->rdb_pipe_read);
+            zfree(g_pserver->rdb_pipe_conns);
+            close(args->safe_to_exit_pipe);
+            g_pserver->rdb_pipe_conns = NULL;
+            g_pserver->rdb_pipe_numconns = 0;
+            g_pserver->rdb_pipe_numconns_writing = 0;
+            args->rsi.~rdbSaveInfo();
+            zfree(args);
+            closeChildInfoPipe();
+            return C_ERR;
+        }
+        pthread_attr_destroy(&tattr);
+        g_pserver->child_type = CHILD_TYPE_RDB;
+
+        serverLog(LL_NOTICE,"Background RDB transfer started");
+        g_pserver->rdb_save_time_start = time(NULL);
+        serverAssert(!g_pserver->rdbThreadVars.fRdbThreadActive);
+        g_pserver->rdbThreadVars.rdb_child_thread = child;
+        g_pserver->rdbThreadVars.fRdbThreadActive = true;
+        g_pserver->rdb_child_type = RDB_CHILD_TYPE_SOCKET;
+        aePostFunction(g_pserver->rgthreadvar[IDX_EVENT_LOOP_MAIN].el, []{
+            if (aeCreateFileEvent(serverTL->el, g_pserver->rdb_pipe_read, AE_READABLE, rdbPipeReadHandler, nullptr) == AE_ERR) {
+                serverPanic("Unrecoverable error creating server.rdb_pipe_read file event.");
+            }
+        });
+    }
 
     return C_OK; /* Unreached. */
 }

src/server.cpp

@@ -5172,7 +5172,7 @@ int prepareForShutdown(int flags) {
          * to unlink file actully) in background thread.
          * The temp rdb file fd may won't be closed when redis exits quickly,
          * but OS will close this fd when process exits. */
-        rdbRemoveTempFile(g_pserver->child_pid, 0);
+        rdbRemoveTempFile(g_pserver->rdbThreadVars.tmpfileNum, 0);
     }
 
     /* Kill module child if there is one. */

src/server.h

@@ -1209,6 +1209,7 @@ public:
     bool FSnapshot() const { return m_spdbSnapshotHOLDER != nullptr; }
 
     std::unique_ptr<const StorageCache> CloneStorageCache() { return std::unique_ptr<const StorageCache>(m_spstorage->clone()); }
+    std::shared_ptr<StorageCache> getStorageCache() { return m_spstorage; }
     void bulkDirectStorageInsert(char **rgKeys, size_t *rgcbKeys, char **rgVals, size_t *rgcbVals, size_t celem);
 
     dict_iter find_cached_threadsafe(const char *key) const;
@@ -1370,6 +1371,7 @@ struct redisDb : public redisDbPersistentDataSnapshot
     using redisDbPersistentData::FRehashing;
     using redisDbPersistentData::FTrackingChanges;
     using redisDbPersistentData::CloneStorageCache;
+    using redisDbPersistentData::getStorageCache;
     using redisDbPersistentData::bulkDirectStorageInsert;
 
 public:
@@ -2729,9 +2731,13 @@ struct redisServer {
     long long repl_batch_offStart = -1;
     long long repl_batch_idxStart = -1;
 
+    long long rand_total_threshold;
+
     int config_soft_shutdown = false;
     bool soft_shutdown = false;
 
+    int flash_disable_key_cache = false;
+
     /* Lock Contention Ring Buffer */
     static const size_t s_lockContentionSamples = 64;
     uint16_t rglockSamples[s_lockContentionSamples];

src/storage/rocksdb.cpp

@@ -3,6 +3,8 @@
 #include <sstream>
 #include <mutex>
 #include <unistd.h>
+#include "../server.h"
+#include "../cluster.h"
 #include "rocksdbfactor_internal.h"
 
 static const char keyprefix[] = INTERNAL_KEY_PREFIX;
@@ -20,6 +22,17 @@ bool FInternalKey(const char *key, size_t cch)
     return false;
 }
 
+std::string getPrefix(unsigned int hashslot)
+{
+    char *hash_char = (char *)&hashslot;
+    return std::string(hash_char + (sizeof(unsigned int) - 2), 2);
+}
+
+std::string prefixKey(const char *key, size_t cchKey)
+{
+    return FInternalKey(key, cchKey) ? std::string(key, cchKey) : getPrefix(keyHashSlot(key, cchKey)) + std::string(key, cchKey);
+}
+
 RocksDBStorageProvider::RocksDBStorageProvider(RocksDBStorageFactory *pfactory, std::shared_ptr<rocksdb::DB> &spdb, std::shared_ptr<rocksdb::ColumnFamilyHandle> &spcolfam, const rocksdb::Snapshot *psnapshot, size_t count)
     : m_pfactory(pfactory), m_spdb(spdb), m_psnapshot(psnapshot), m_spcolfamily(spcolfam), m_count(count)
 {
@@ -32,10 +45,11 @@ void RocksDBStorageProvider::insert(const char *key, size_t cchKey, void *data,
 {
     rocksdb::Status status;
     std::unique_lock<fastlock> l(m_lock);
+    std::string prefixed_key = prefixKey(key, cchKey);
     if (m_spbatch != nullptr)
-        status = m_spbatch->Put(m_spcolfamily.get(), rocksdb::Slice(key, cchKey), rocksdb::Slice((const char*)data, cb));
+        status = m_spbatch->Put(m_spcolfamily.get(), rocksdb::Slice(prefixed_key), rocksdb::Slice((const char*)data, cb));
     else
-        status = m_spdb->Put(WriteOptions(), m_spcolfamily.get(), rocksdb::Slice(key, cchKey), rocksdb::Slice((const char*)data, cb));
+        status = m_spdb->Put(WriteOptions(), m_spcolfamily.get(), rocksdb::Slice(prefixed_key), rocksdb::Slice((const char*)data, cb));
     if (!status.ok())
         throw status.ToString();
 
@@ -59,7 +73,8 @@ void RocksDBStorageProvider::bulkInsert(char **rgkeys, size_t *rgcbkeys, char **
         // Insert rows into the SST file, note that inserted keys must be 
         // strictly increasing (based on options.comparator)
         for (size_t ielem = 0; ielem < celem; ++ielem) {
-            s = sst_file_writer.Put(rocksdb::Slice(rgkeys[ielem], rgcbkeys[ielem]), rocksdb::Slice(rgvals[ielem], rgcbvals[ielem]));
+            std::string prefixed_key = prefixKey(rgkeys[ielem], rgcbkeys[ielem]);
+            s = sst_file_writer.Put(rocksdb::Slice(prefixed_key), rocksdb::Slice(rgvals[ielem], rgcbvals[ielem]));
             if (!s.ok()) {
                 unlink(file_path.c_str());
                 goto LFallback;
@@ -87,7 +102,8 @@ void RocksDBStorageProvider::bulkInsert(char **rgkeys, size_t *rgcbkeys, char **
     LFallback:
         auto spbatch = std::make_unique<rocksdb::WriteBatch>();
         for (size_t ielem = 0; ielem < celem; ++ielem) {
-            spbatch->Put(m_spcolfamily.get(), rocksdb::Slice(rgkeys[ielem], rgcbkeys[ielem]), rocksdb::Slice(rgvals[ielem], rgcbvals[ielem]));
+            std::string prefixed_key = prefixKey(rgkeys[ielem], rgcbkeys[ielem]);
+            spbatch->Put(m_spcolfamily.get(), rocksdb::Slice(prefixed_key), rocksdb::Slice(rgvals[ielem], rgcbvals[ielem]));
         }
         m_spdb->Write(WriteOptions(), spbatch.get());
     }
@@ -100,15 +116,16 @@ bool RocksDBStorageProvider::erase(const char *key, size_t cchKey)
 {
     rocksdb::Status status;
     std::unique_lock<fastlock> l(m_lock);
-    if (!FKeyExists(key, cchKey))
+    std::string prefixed_key = prefixKey(key, cchKey);
+    if (!FKeyExists(prefixed_key))
         return false;
     if (m_spbatch != nullptr)
     {
-        status = m_spbatch->Delete(m_spcolfamily.get(), rocksdb::Slice(key, cchKey));
+        status = m_spbatch->Delete(m_spcolfamily.get(), rocksdb::Slice(prefixed_key));
     }
     else
     {
-        status = m_spdb->Delete(WriteOptions(), m_spcolfamily.get(), rocksdb::Slice(key, cchKey));
+        status = m_spdb->Delete(WriteOptions(), m_spcolfamily.get(), rocksdb::Slice(prefixed_key));
     }
     if (status.ok())
         --m_count;
@@ -118,7 +135,8 @@ bool RocksDBStorageProvider::erase(const char *key, size_t cchKey)
 void RocksDBStorageProvider::retrieve(const char *key, size_t cchKey, callbackSingle fn) const
 {
     rocksdb::PinnableSlice slice;
-    auto status = m_spdb->Get(ReadOptions(), m_spcolfamily.get(), rocksdb::Slice(key, cchKey), &slice);
+    std::string prefixed_key = prefixKey(key, cchKey);
+    auto status = m_spdb->Get(ReadOptions(), m_spcolfamily.get(), rocksdb::Slice(prefixed_key), &slice);
     if (status.ok())
         fn(key, cchKey, slice.data(), slice.size());
 }
@@ -154,7 +172,7 @@ bool RocksDBStorageProvider::enumerate(callback fn) const
         if (FInternalKey(it->key().data(), it->key().size()))
             continue;
         ++count;
-        bool fContinue = fn(it->key().data(), it->key().size(), it->value().data(), it->value().size());
+        bool fContinue = fn(it->key().data()+2, it->key().size()-2, it->value().data(), it->value().size());
         if (!fContinue)
             break;
     }
@@ -168,6 +186,31 @@ bool RocksDBStorageProvider::enumerate(callback fn) const
     return !it->Valid();
 }
 
+bool RocksDBStorageProvider::enumerate_hashslot(callback fn, unsigned int hashslot) const
+{
+    std::string prefix = getPrefix(hashslot);
+    std::unique_ptr<rocksdb::Iterator> it = std::unique_ptr<rocksdb::Iterator>(m_spdb->NewIterator(ReadOptions(), m_spcolfamily.get()));
+    size_t count = 0;
+    for (it->Seek(prefix.c_str()); it->Valid(); it->Next()) {
+        if (FInternalKey(it->key().data(), it->key().size()))
+            continue;
+        if (strncmp(it->key().data(),prefix.c_str(),2) != 0)
+            break;
+        ++count;
+        bool fContinue = fn(it->key().data()+2, it->key().size()-2, it->value().data(), it->value().size());
+        if (!fContinue)
+            break;
+    }
+    bool full_iter = !it->Valid() || (strncmp(it->key().data(),prefix.c_str(),2) != 0);
+    if (full_iter && count != g_pserver->cluster->slots_keys_count[hashslot])
+    {
+        printf("WARNING: rocksdb hashslot count mismatch");
+    }
+    assert(!full_iter || count == g_pserver->cluster->slots_keys_count[hashslot]);
+    assert(it->status().ok()); // Check for any errors found during the scan
+    return full_iter;
+}
+
 const IStorage *RocksDBStorageProvider::clone() const
 {
     std::unique_lock<fastlock> l(m_lock);
@@ -227,10 +270,10 @@ void RocksDBStorageProvider::flush()
     m_spdb->SyncWAL();
 }
 
-bool RocksDBStorageProvider::FKeyExists(const char *key, size_t cch) const
+bool RocksDBStorageProvider::FKeyExists(std::string& key) const
 {
     rocksdb::PinnableSlice slice;
     if (m_spbatch)
-        return m_spbatch->GetFromBatchAndDB(m_spdb.get(), ReadOptions(), m_spcolfamily.get(), rocksdb::Slice(key, cch), &slice).ok();
-    return m_spdb->Get(ReadOptions(), m_spcolfamily.get(), rocksdb::Slice(key, cch), &slice).ok();
+        return m_spbatch->GetFromBatchAndDB(m_spdb.get(), ReadOptions(), m_spcolfamily.get(), rocksdb::Slice(key), &slice).ok();
+    return m_spdb->Get(ReadOptions(), m_spcolfamily.get(), rocksdb::Slice(key), &slice).ok();
 }

src/storage/rocksdb.h

@@ -32,6 +32,7 @@ public:
     virtual void retrieve(const char *key, size_t cchKey, callbackSingle fn) const override;
     virtual size_t clear() override;
     virtual bool enumerate(callback fn) const override;
+    virtual bool enumerate_hashslot(callback fn, unsigned int hashslot) const override;
 
     virtual const IStorage *clone() const override;
 
@@ -48,7 +49,7 @@ public:
     size_t count() const override;
 
 protected:
-    bool FKeyExists(const char *key, size_t cchKey) const;
+    bool FKeyExists(std::string&) const;
 
     const rocksdb::ReadOptions &ReadOptions() const { return m_readOptionsTemplate; }
     rocksdb::WriteOptions WriteOptions() const;

src/storage/rocksdbfactory.cpp

@@ -71,6 +71,7 @@ RocksDBStorageFactory::RocksDBStorageFactory(const char *dbfile, int dbnum, cons
     rocksdb::DB *db = nullptr;
 
     auto options = RocksDbOptions();
+    options.prefix_extractor.reset(rocksdb::NewFixedPrefixTransform(2));
 
     for (int idb = 0; idb < dbnum; ++idb)
     {
@@ -186,7 +187,7 @@ IStorage *RocksDBStorageFactory::create(int db, key_load_iterator iter, void *pr
                 printf("\tDatabase %d was not shutdown cleanly, recomputing metrics\n", db);
             fFirstRealKey = false;
             if (iter != nullptr)
-                iter(it->key().data(), it->key().size(), privdata);
+                iter(it->key().data()+2, it->key().size()-2, privdata);
             ++count;
         }
     }

src/storage/teststorageprovider.cpp

@@ -73,7 +73,22 @@ bool TestStorageProvider::enumerate(callback fn) const
     }
     return fAll;
 }
-    
+
+bool TestStorageProvider::enumerate_hashslot(callback fn, unsigned int hashslot) const
+{
+    bool fAll = true;
+    for (auto &pair : m_map)
+    {
+        if (keyHashSlot(pair.first.data(), pair.first.size()) == hashslot)
+            if (!fn(pair.first.data(), pair.first.size(), pair.second.data(), pair.second.size()))
+            {
+                fAll = false;
+                break;
+            }
+    }
+    return fAll;
+}
+      
 size_t TestStorageProvider::count() const
 {
     return m_map.size();

src/storage/teststorageprovider.h

@@ -24,6 +24,7 @@ public:
     virtual void retrieve(const char *key, size_t cchKey, callbackSingle fn) const override;
     virtual size_t clear() override;
     virtual bool enumerate(callback fn) const override;
+    virtual bool enumerate_hashslot(callback fn, unsigned int hashslot) const override;
     virtual size_t count() const override;
 
     virtual void flush() override;

src/t_hash.cpp

@@ -729,6 +729,10 @@ void hincrbyfloatCommand(client *c) {
     unsigned int vlen;
 
     if (getLongDoubleFromObjectOrReply(c,c->argv[3],&incr,NULL) != C_OK) return;
+    if (isnan(incr) || isinf(incr)) {
+        addReplyError(c,"value is NaN or Infinity");
+        return;
+    }
     if ((o = hashTypeLookupWriteOrCreate(c,c->argv[1])) == NULL) return;
     if (hashTypeGetValue(o,szFromObj(c->argv[2]),&vstr,&vlen,&ll) == C_OK) {
         if (vstr) {
@@ -1062,6 +1066,8 @@ void hrandfieldWithCountCommand(client *c, long l, int withvalues) {
                 addReplyBulkCBuffer(c, key, sdslen(key));
                 if (withvalues)
                     addReplyBulkCBuffer(c, value, sdslen(value));
+                if (c->flags & CLIENT_CLOSE_ASAP)
+                    break;
             }
         } else if (hash->encoding == OBJ_ENCODING_ZIPLIST) {
             ziplistEntry *keys, *vals = NULL;
@@ -1075,6 +1081,8 @@ void hrandfieldWithCountCommand(client *c, long l, int withvalues) {
                 count -= sample_count;
                 ziplistRandomPairs((unsigned char*)ptrFromObj(hash), sample_count, keys, vals);
                 harndfieldReplyWithZiplist(c, sample_count, keys, vals);
+                if (c->flags & CLIENT_CLOSE_ASAP)
+                    break;
             }
             zfree(keys);
             zfree(vals);
@@ -1221,13 +1229,13 @@ void hrandfieldCommand(client *c) {
     ziplistEntry ele;
 
     if (c->argc >= 3) {
-        if (getLongFromObjectOrReply(c,c->argv[2],&l,NULL) != C_OK) return;
+        if (getRangeLongFromObjectOrReply(c,c->argv[2],-LONG_MAX,LONG_MAX,&l,NULL) != C_OK) return;
         if (c->argc > 4 || (c->argc == 4 && strcasecmp(szFromObj(c->argv[3]),"withvalues"))) {
             addReplyErrorObject(c,shared.syntaxerr);
             return;
         } else if (c->argc == 4) {
             withvalues = 1;
-            if (l < LONG_MIN/2 || l > LONG_MAX/2) {
+            if (l < -g_pserver->rand_total_threshold || l > g_pserver->rand_total_threshold) {
                 addReplyError(c,"value is out of range");
                 return;
             }

src/t_set.cpp

@@ -672,7 +672,11 @@ void srandmemberWithCountCommand(client *c) {
 
     dict *d;
 
-    if (getLongFromObjectOrReply(c,c->argv[2],&l,NULL) != C_OK) return;
+    if (getRangeLongFromObjectOrReply(c,c->argv[2],-LONG_MAX,LONG_MAX,&l,NULL) != C_OK) return;
+    if (l < -g_pserver->rand_total_threshold || l > g_pserver->rand_total_threshold) {
+        addReplyError(c,"value is out of range");
+        return;
+    }
     if (l >= 0) {
         count = (unsigned long) l;
     } else {
@@ -706,6 +710,8 @@ void srandmemberWithCountCommand(client *c) {
             } else {
                 addReplyBulkCBuffer(c,ele,sdslen(ele));
             }
+            if (c->flags & CLIENT_CLOSE_ASAP)
+                break;
         }
         return;
     }

src/t_zset.cpp

@@ -4054,6 +4054,8 @@ void zrandmemberWithCountCommand(client *c, long l, int withscores) {
                 addReplyBulkCBuffer(c, key, sdslen(key));
                 if (withscores)
                     addReplyDouble(c, *(double*)dictGetVal(de));
+                if (c->flags & CLIENT_CLOSE_ASAP)
+                    break;
             }
         } else if (zsetobj->encoding == OBJ_ENCODING_ZIPLIST) {
             ziplistEntry *keys, *vals = NULL;
@@ -4067,6 +4069,8 @@ void zrandmemberWithCountCommand(client *c, long l, int withscores) {
                 count -= sample_count;
                 ziplistRandomPairs((unsigned char*)ptrFromObj(zsetobj), sample_count, keys, vals);
                 zarndmemberReplyWithZiplist(c, sample_count, keys, vals);
+                if (c->flags & CLIENT_CLOSE_ASAP)
+                    break;
             }
             zfree(keys);
             zfree(vals);
@@ -4210,13 +4214,13 @@ void zrandmemberCommand(client *c) {
     ziplistEntry ele;
 
     if (c->argc >= 3) {
-        if (getLongFromObjectOrReply(c,c->argv[2],&l,NULL) != C_OK) return;
+        if (getRangeLongFromObjectOrReply(c,c->argv[2],-LONG_MAX,LONG_MAX,&l,NULL) != C_OK) return;
         if (c->argc > 4 || (c->argc == 4 && strcasecmp(szFromObj(c->argv[3]),"withscores"))) {
             addReplyErrorObject(c,shared.syntaxerr);
             return;
         } else if (c->argc == 4) {
             withscores = 1;
-            if (l < LONG_MIN/2 || l > LONG_MAX/2) {
+            if (l < -g_pserver->rand_total_threshold || l > g_pserver->rand_total_threshold) {
                 addReplyError(c,"value is out of range");
                 return;
             }

tests/instances.tcl

@@ -21,6 +21,7 @@ set ::tls 0
 set ::pause_on_error 0
 set ::dont_clean 0
 set ::simulate_error 0
+set ::flash 0
 set ::failed 0
 set ::sentinel_instances {}
 set ::redis_instances {}
@@ -81,6 +82,10 @@ proc spawn_instance {type base_port count {conf {}} {base_conf_file ""}} {
             set cfg [open $cfgfile w]
         }
 
+        if {$::flash} {
+            puts $cfg "storage-provider flash ./flash_$base_port"
+        }
+
         if {$::tls} {
             puts $cfg "tls-port $port"
             puts $cfg "tls-replication yes"
@@ -266,6 +271,8 @@ proc parse_options {} {
             set val2 [lindex $::argv [expr $j+2]]
             dict set ::global_config $val $val2
             incr j 2
+        } elseif {$opt eq {--flash}} {
+            set ::flash 1
         } elseif {$opt eq "--help"} {
             puts "--single <pattern>      Only runs tests specified by pattern."
             puts "--dont-clean            Keep log files on exit."
@@ -275,6 +282,7 @@ proc parse_options {} {
             puts "--tls                   Run tests in TLS mode."
             puts "--host <host>           Use hostname instead of 127.0.0.1."
             puts "--config <k> <v>        Extra config argument(s)."
+            puts "--flash                 Run the whole suite with flash enabled"
             puts "--help                  Shows this help."
             exit 0
         } else {
@@ -301,11 +309,15 @@ proc pause_on_error {} {
             break
         } elseif {$cmd eq {show-keydb-logs}} {
             set count 10
+            set instance {}
             if {[lindex $argv 1] ne {}} {set count [lindex $argv 1]}
+            if {[lindex $argv 2] ne {}} {set instance [lindex $argv 2]}
             foreach_redis_id id {
-                puts "=== KeyDB $id ===="
-                puts [exec tail -$count redis_$id/log.txt]
-                puts "---------------------\n"
+                if {$instance eq $id || $instance eq {}} {
+                    puts "=== KeyDB $id ===="
+                    puts [exec tail -$count redis_$id/log.txt]
+                    puts "---------------------\n"
+                }
             }
         } elseif {$cmd eq {show-sentinel-logs}} {
             set count 10
@@ -350,7 +362,7 @@ proc pause_on_error {} {
         } elseif {$cmd eq {help}} {
             puts "ls                     List Sentinel and KeyDB instances."
             puts "show-sentinel-logs \[N\] Show latest N lines of logs."
-            puts "show-keydb-logs \[N\]    Show latest N lines of logs."
+            puts "show-keydb-logs \[N\] \[id\]    Show latest N lines of logs of server id."
             puts "S <id> cmd ... arg     Call command in Sentinel <id>."
             puts "R <id> cmd ... arg     Call command in KeyDB <id>."
             puts "SI <id> <field>        Show Sentinel <id> INFO <field>."

tests/unit/obuf-limits.tcl

@@ -182,4 +182,12 @@ start_server {tags {"obuf-limits"} overrides { server-threads 1 }} {
         assert_equal "v2" [r get k2]
         assert_equal "v3" [r get k3]
     }
+
+    test "Obuf limit, HRANDFIELD with huge count stopped mid-run" {
+        r config set client-output-buffer-limit {normal 1000000 0 0}
+        r hset myhash a b
+        catch {r hrandfield myhash -999999999} e
+        assert_match "*I/O error*" $e
+        reconnect
+    }
 }

tests/unit/type/hash.tcl

@@ -71,6 +71,8 @@ start_server {tags {"hash"}} {
     test "HRANDFIELD count overflow" {
         r hmset myhash a 1
         assert_error {*value is out of range*} {r hrandfield myhash -9223372036854770000 withvalues}
+        assert_error {*value is out of range*} {r hrandfield myhash -9223372036854775808 withvalues}
+        assert_error {*value is out of range*} {r hrandfield myhash -9223372036854775808}
     } {}
 
     test "HRANDFIELD with <count> against non existing key" {
@@ -831,4 +833,8 @@ start_server {tags {"hash"}} {
         set _ $k
     } {ZIP_INT_8B 127 ZIP_INT_16B 32767 ZIP_INT_32B 2147483647 ZIP_INT_64B 9223372036854775808 ZIP_INT_IMM_MIN 0 ZIP_INT_IMM_MAX 12}
 
+    test {HINCRBYFLOAT does not allow NaN or Infinity} {
+        assert_error "*value is NaN or Infinity*" {r hincrbyfloat hfoo field +inf}
+        assert_equal 0 [r exists hfoo]
+    }
 }

tests/unit/type/set.tcl

@@ -588,6 +588,11 @@ start_server {
         r srandmember nonexisting_key 100
     } {}
 
+    test "SRANDMEMBER count overflow" {
+        r sadd myset a
+        assert_error {*value is out of range*} {r srandmember myset -9223372036854775808}
+    } {}
+
     # Make sure we can distinguish between an empty array and a null response
     r readraw 1
 

tests/unit/type/zset.tcl

@@ -1717,6 +1717,8 @@ start_server {tags {"zset"}} {
     test "ZRANDMEMBER count overflow" {
         r zadd myzset 0 a
         assert_error {*value is out of range*} {r zrandmember myzset -9223372036854770000 withscores}
+        assert_error {*value is out of range*} {r zrandmember myzset -9223372036854775808 withscores}
+        assert_error {*value is out of range*} {r zrandmember myzset -9223372036854775808}
     } {}
 
     # Make sure we can distinguish between an empty array and a null response

utils/create-cluster/create-cluster

@@ -26,7 +26,11 @@ then
     while [ $((PORT < ENDPORT)) != "0" ]; do
         PORT=$((PORT+1))
         echo "Starting $PORT"
-        $BIN_PATH/keydb-server --port $PORT  --protected-mode $PROTECTED_MODE --cluster-enabled yes --cluster-config-file nodes-${PORT}.conf --cluster-node-timeout $TIMEOUT --appendonly yes --appendfilename appendonly-${PORT}.aof --dbfilename dump-${PORT}.rdb --logfile ${PORT}.log --daemonize yes ${ADDITIONAL_OPTIONS}
+        if [ "$2" == "flash" ]
+        then
+            ADDITIONAL_OPTIONS="--save \"\" \"\" \"\" --semi-ordered-set-bucket-size 8 --client-output-buffer-limit replica 1 1 0 --maxmemory 100000000 --storage-provider flash ./$PORT.flash"
+        fi
+        $BIN_PATH/keydb-server --server-threads 4 --port $PORT  --protected-mode $PROTECTED_MODE --cluster-enabled yes --cluster-config-file nodes-${PORT}.conf --cluster-node-timeout $TIMEOUT --dbfilename dump-${PORT}.rdb --logfile ${PORT}.log --daemonize yes ${ADDITIONAL_OPTIONS}
     done
     exit 0
 fi
@@ -97,6 +101,8 @@ then
     rm -rf appendonly*.aof
     rm -rf dump*.rdb
     rm -rf nodes*.conf
+    rm -rf *.flash
+    rm -rf temp*.rdb
     exit 0
 fi