From 34f7c653d54e428bed3a6fc686f9612ccbbdd348 Mon Sep 17 00:00:00 2001
From: antirez <antirez@gmail.com>
Date: Tue, 3 Mar 2020 14:58:11 +0100
Subject: [PATCH] Remove RDB files used for replication in persistence-less
 instances.

Former-commit-id: b323645227a3e2cc5928e649586221aba508b10d
---
 src/replication.cpp | 49 ++++++++++++++++++++++++++++++++++++++++++++-
 src/server.cpp      |  8 ++++++++
 src/server.h        |  1 +
 3 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/replication.cpp b/src/replication.cpp
index 1edbb9b06..3d0cafc56 100644
--- a/src/replication.cpp
+++ b/src/replication.cpp
@@ -53,6 +53,11 @@ void putSlaveOnline(client *replica);
 int cancelReplicationHandshake(redisMaster *mi);
 static void propagateMasterStaleKeys();
 
+/* We take a global flag to remember if this instance generated an RDB
+ * because of replication, so that we can remove the RDB file in case
+ * the instance is configured to have no persistence. */
+int RDBGeneratedByReplication = 0;
+
 /* --------------------------- Utility functions ---------------------------- */
 
 /* Return the pointer to a string representing the replica ip:listening_port
@@ -789,6 +794,10 @@ int startBgsaveForReplication(int mincapa) {
         retval = C_ERR;
     }
 
+    /* If we succeeded to start a BGSAVE with disk target, let's remember
+     * this fact, so that we can later delete the file if needed. */
+    if (retval == C_OK && !socket_target) RDBGeneratedByReplication = 1;
+
     /* If we failed to BGSAVE, remove the slaves waiting for a full
      * resynchronization from the list of slaves, inform them with
      * an error about what happened, close the connection ASAP. */
@@ -1132,6 +1141,36 @@ void putSlaveOnline(client *replica) {
     }
 }
 
+/* We call this function periodically to remove an RDB file that was
+ * generated because of replication, in an instance that is otherwise
+ * without any persistence. We don't want instances without persistence
+ * to take RDB files around, this violates certain policies in certain
+ * environments. */
+void removeRDBUsedToSyncReplicas(void) {
+    if (allPersistenceDisabled() && RDBGeneratedByReplication) {
+        client *slave;
+        listNode *ln;
+        listIter li;
+
+        int delrdb = 1;
+        listRewind(server.slaves,&li);
+        while((ln = listNext(&li))) {
+            slave = ln->value;
+            if (slave->replstate == SLAVE_STATE_WAIT_BGSAVE_START ||
+                slave->replstate == SLAVE_STATE_WAIT_BGSAVE_END ||
+                slave->replstate == SLAVE_STATE_SEND_BULK)
+            {
+                delrdb = 0;
+                break; /* No need to check the other replicas. */
+            }
+        }
+        if (delrdb) {
+            RDBGeneratedByReplication = 0;
+            unlink(server.rdb_filename);
+        }
+    }
+}
+
 void sendBulkToSlave(connection *conn) {
     client *replica = (client*)connGetPrivateData(conn);
     serverAssert(FCorrectThread(replica));
@@ -1144,7 +1183,8 @@ void sendBulkToSlave(connection *conn) {
     if (replica->replpreamble) {
         nwritten = connWrite(conn,replica->replpreamble,sdslen(replica->replpreamble));
         if (nwritten == -1) {
-            serverLog(LL_VERBOSE,"Write error sending RDB preamble to replica: %s",
+            serverLog(LL_VERBOSE,
+                "Write error sending RDB preamble to replica: %s",
                 connGetLastError(conn));
             freeClient(replica);
             return;
@@ -1985,12 +2025,14 @@ void readSyncBulkPayload(connection *conn) {
                 "Failed trying to load the MASTER synchronization "
                 "DB from disk");
             cancelReplicationHandshake(mi);
+            if (allPersistenceDisabled()) unlink(g_pserver->rdb_filename);
             /* Note that there's no point in restarting the AOF on sync failure,
                it'll be restarted when sync succeeds or replica promoted. */
             return;
         }
 
         /* Cleanup. */
+        if (allPersistenceDisabled()) unlink(server.rdb_filename);
         if (fUpdate)
             unlink(mi->repl_transfer_tmpfile);
         zfree(mi->repl_transfer_tmpfile);
@@ -3684,6 +3726,11 @@ void replicationCron(void) {
 
     propagateMasterStaleKeys();
 
+
+    /* Remove the RDB file used for replication if Redis is not running
+     * with any persistence. */
+    removeRDBUsedToSyncReplicas();
+
     /* Refresh the number of slaves with lag <= min-slaves-max-lag. */
     refreshGoodSlavesCount();
     replication_cron_loops++; /* Incremented with frequency 1 HZ. */
diff --git a/src/server.cpp b/src/server.cpp
index 874d391c6..30f4f4099 100644
--- a/src/server.cpp
+++ b/src/server.cpp
@@ -1483,12 +1483,20 @@ void updateDictResizePolicy(void) {
         dictDisableResize();
 }
 
+/* Return true if there are no active children processes doing RDB saving,
+ * AOF rewriting, or some side process spawned by a loaded module. */
 int hasActiveChildProcess() {
     return g_pserver->rdb_child_pid != -1 ||
            g_pserver->aof_child_pid != -1 ||
            g_pserver->module_child_pid != -1;
 }
 
+/* Return true if this instance has persistence completely turned off:
+ * both RDB and AOF are disabled. */
+int allPersistenceDisabled(void) {
+    return server.saveparamslen == 0 && server.aof_state == AOF_OFF;
+}
+
 /* ======================= Cron: called every 100 ms ======================== */
 
 /* Add a sample to the operations per second array of samples. */
diff --git a/src/server.h b/src/server.h
index b8a11e888..b9bd9b7c4 100644
--- a/src/server.h
+++ b/src/server.h
@@ -2357,6 +2357,7 @@ void loadingProgress(off_t pos);
 void stopLoading(int success);
 void startSaving(int rdbflags);
 void stopSaving(int success);
+int allPersistenceDisabled(void);
 
 #define DISK_ERROR_TYPE_AOF 1       /* Don't accept writes: AOF errors. */
 #define DISK_ERROR_TYPE_RDB 2       /* Don't accept writes: RDB errors. */