MDEV-39126: Feedback plugin to use /etc/os-release

Update the feedback plugin to use /etc/os-release for OS discovery,
as /etc/lsb-release is becoming obsolete. Maintained backwards
compatibility by keeping the old checks as fallbacks.

Signed-off-by: Anway Durge <durgeanway@gmail.com>

Author: itzanway

mysql-test/suite/feedback/r/os_release.result

@@ -0,0 +1,7 @@
+#
+# MDEV-39126: Feedback plugin to use /etc/os-release
+#
+SELECT variable_name, variable_value FROM information_schema.feedback WHERE variable_name = 'os';
+variable_name	variable_value
+os	MASKED_OS_INFO
+

mysql-test/suite/feedback/t/os_release.test

@@ -0,0 +1,9 @@
+--source include/have_feedback.inc
+
+--echo #
+--echo # MDEV-39126: Feedback plugin to use /etc/os-release
+--echo #
+
+--replace_column 2 MASKED_OS_INFO
+SELECT variable_name, variable_value FROM information_schema.feedback WHERE variable_name = 'os';  
+    

mysql-test/suite/galera/r/MDEV-39011.result

@@ -0,0 +1,21 @@
+connection node_2;
+connection node_1;
+connect node_3, 127.0.0.1, root, , test, $NODE_MYPORT_3;
+connection node_3;
+RESET MASTER;
+connection node_2;
+SET GLOBAL DEBUG_DBUG = "+d,delay_sql_thread_after_release_run_lock";
+START SLAVE;
+SET DEBUG_SYNC = "now WAIT_FOR sql_thread_run_lock_released";
+SET GLOBAL DEBUG_DBUG = "+d,wsrep_async_slave_node_dropped_error";
+SET DEBUG_SYNC = "now SIGNAL sql_thread_continue";
+SET DEBUG_SYNC = "now WAIT_FOR sql_thread_run_lock_released";
+SET GLOBAL DEBUG_DBUG = "-d,wsrep_async_slave_node_dropped_error";
+SET DEBUG_SYNC = "now SIGNAL sql_thread_continue";
+SET DEBUG_SYNC = "RESET";
+SET GLOBAL DEBUG_DBUG = "";
+STOP SLAVE;
+RESET SLAVE;
+connection node_3;
+RESET MASTER;
+disconnect node_3;

mysql-test/suite/galera/t/MDEV-39011.cnf

@@ -0,0 +1,7 @@
+!include ../galera_2nodes_as_slave.cnf
+
+[mysqld.1]
+wsrep_restart_slave=1
+
+[mysqld.2]
+wsrep_restart_slave=1

mysql-test/suite/galera/t/MDEV-39011.test

@@ -0,0 +1,48 @@
+#
+# MDEV-39011: The node running async slave replication hangs in shutdown.
+# The problem was in a leaked THD_count on the error path when
+# wsrep_restart_slave is set.
+#
+
+--source include/galera_cluster.inc
+--source include/have_innodb.inc
+--source include/have_debug.inc
+--source include/have_debug_sync.inc
+
+# node 3 is a native MariaDB server operating as async replication master
+--connect node_3, 127.0.0.1, root, , test, $NODE_MYPORT_3
+--connection node_3
+RESET MASTER;
+
+# nodes 1 and 2 form a Galera cluster, node 2 operates as a slave for the
+# native MariaDB master in node 3
+--connection node_2
+--disable_query_log
+--eval CHANGE MASTER TO MASTER_HOST='127.0.0.1', MASTER_USER='root', MASTER_PORT=$NODE_MYPORT_3;
+--enable_query_log
+
+SET GLOBAL DEBUG_DBUG = "+d,delay_sql_thread_after_release_run_lock";
+START SLAVE;
+# Wait for the async slave thread to get started
+SET DEBUG_SYNC = "now WAIT_FOR sql_thread_run_lock_released";
+# Simulate a node dropped error for Wsrep to go to the restart point
+SET GLOBAL DEBUG_DBUG = "+d,wsrep_async_slave_node_dropped_error";
+SET DEBUG_SYNC = "now SIGNAL sql_thread_continue";
+# Now wait again as the execution should follow the error path and repeat
+SET DEBUG_SYNC = "now WAIT_FOR sql_thread_run_lock_released";
+# Don't follow the error path anymore
+SET GLOBAL DEBUG_DBUG = "-d,wsrep_async_slave_node_dropped_error";
+SET DEBUG_SYNC = "now SIGNAL sql_thread_continue";
+
+# Cleanup before restart
+SET DEBUG_SYNC = "RESET";
+SET GLOBAL DEBUG_DBUG = "";
+STOP SLAVE;
+RESET SLAVE;
+
+# Now try to restart the node: it should not hang in a shutdown
+--source include/restart_mysqld.inc
+
+--connection node_3
+RESET MASTER;
+--disconnect node_3

mysql-test/suite/galera/t/galera_as_slave_replay.cnf

@@ -4,9 +4,7 @@
 binlog-format=row
 
 [mysqld.1]
-wsrep_restart_slave=1
 wsrep-debug=1
 
 [mysqld.2]
-wsrep_restart_slave=1
 wsrep-debug=1

plugin/feedback/utils.cc

@@ -266,27 +266,59 @@ int prepare_linux_info()
 
 #ifdef TARGET_OS_LINUX
   /*
-    let's try to find what linux distribution it is
-    we read *[-_]{release,version} file in /etc.
+    Let's try to find what Linux distribution it is.
+    We first check for the modern and portable /etc/os-release:
 
-    Either it will be /etc/lsb-release, such as
+      ==> /etc/os-release <==
+      PRETTY_NAME="Ubuntu 22.04.2 LTS"
+
+    If not found, we fall back to /etc/lsb-release:
 
       ==> /etc/lsb-release <==
-      DISTRIB_ID=Ubuntu
-      DISTRIB_RELEASE=8.04
-      DISTRIB_CODENAME=hardy
       DISTRIB_DESCRIPTION="Ubuntu 8.04.4 LTS"
 
-   Or a one-liner with the description (/etc/SuSE-release has more
-   than one line, but the description is the first, so it can be
-   treated as a one-liner).
+    Or a one-liner with the description (/etc/SuSE-release has more
+    than one line, but the description is the first, so it can be
+    treated as a one-liner).
 
-   We'll read lsb-release first, and if it's not found will search
-   for other files (*-version *-release *_version *_release)
-*/
+    We'll read os-release first, then lsb-release, and if neither is found
+    we will search for other files (*-version *-release *_version *_release).
+  */
   int fd;
   have_distribution= false;
-  if ((fd= my_open("/etc/lsb-release", O_RDONLY, MYF(0))) != -1)
+
+  /* 1. First try the modern portable way: /etc/os-release */
+  if ((fd= my_open("/etc/os-release", O_RDONLY, MYF(0))) != -1)
+  {
+    size_t len= my_read(fd, (uchar*)distribution, sizeof(distribution)-1, MYF(0));
+    my_close(fd, MYF(0));
+    if (len != (size_t)-1)
+    {
+      distribution[len]= 0; // safety
+      char *found= strstr(distribution, "PRETTY_NAME=");
+      if (found)
+      {
+        have_distribution= true;
+        char *end= strstr(found, "\n");
+        if (end == NULL)
+          end= distribution + len;
+        found+= 12; // Length of "PRETTY_NAME="
+
+        if (*found == '"' && end[-1] == '"')
+        {
+          found++;
+          end--;
+        }
+        *end= 0;
+
+        char *to= strmov(distribution, "os-release: ");
+        memmove(to, found, end - found + 1);
+      }
+    }
+  }
+
+  /* 2. Fallback to older /etc/lsb-release if os-release is not found */
+  if (!have_distribution && (fd= my_open("/etc/lsb-release", O_RDONLY, MYF(0))) != -1)
   {
     /* Cool, LSB-compliant distribution! */
     size_t len= my_read(fd, (uchar*)distribution, sizeof(distribution)-1, MYF(0));
@@ -301,7 +333,7 @@ int prepare_linux_info()
         char *end= strstr(found, "\n");
         if (end == NULL)
           end= distribution + len;
-        found+= 20;
+        found+= 20; // Length of "DISTRIB_DESCRIPTION="
 
         if (*found == '"' && end[-1] == '"')
         {
@@ -316,6 +348,7 @@ int prepare_linux_info()
     }
   }
 
+
   /* if not an LSB-compliant distribution */
   for (uint i= 0; !have_distribution && i < array_elements(masks); i++)
   {

sql/mysqld.cc

@@ -6070,11 +6070,7 @@ int mysqld_main(int argc, char **argv)
     my_free(user);
 
 #ifdef WITH_WSREP
-  /* Stop wsrep threads in case they are running. */
-  if (wsrep_running_threads > 0)
-  {
-    wsrep_shutdown_replication();
-  }
+  wsrep_shutdown();
   /* Release threads if they are waiting in WSREP_SYNC_WAIT_UPTO_GTID */
   wsrep_gtid_server.signal_waiters(0, true);
 #endif

sql/slave.cc

@@ -5740,6 +5740,15 @@ pthread_handler_t handle_slave_sql(void *arg)
     DBUG_ASSERT(debug_sync_service);
     DBUG_ASSERT(!debug_sync_set_action(current_thd, STRING_WITH_LEN(act)));
   };);
+#ifdef WITH_WSREP
+  DBUG_EXECUTE_IF("wsrep_async_slave_node_dropped_error",
+    if (WSREP(thd))
+    {
+      wsrep_node_dropped= TRUE;
+      goto err_before_start;
+    }
+  );
+#endif /* WITH_WSREP */
 #endif
 
   rli->parallel.reset();
@@ -5913,8 +5922,8 @@ pthread_handler_t handle_slave_sql(void *arg)
   mysql_mutex_unlock(&rli->data_lock);
 #ifdef WITH_WSREP
   wsrep_open(thd);
-  if (WSREP_ON_)
-    wsrep_wait_ready(thd);
+  if (WSREP_ON_ && !wsrep_wait_ready(thd))
+    goto err;
   if (wsrep_before_command(thd))
   {
     WSREP_WARN("Slave SQL wsrep_before_command() failed");
@@ -6137,21 +6146,25 @@ pthread_handler_t handle_slave_sql(void *arg)
   */
   if (WSREP(thd) && wsrep_node_dropped && wsrep_restart_slave)
   {
-    if (wsrep_ready_get())
-    {
-      WSREP_INFO("Slave error due to node temporarily non-primary"
-                 "SQL slave will continue");
       wsrep_node_dropped= FALSE;
       mysql_mutex_unlock(&rli->run_lock);
-      goto wsrep_restart_point;
-    }
-    else
-    {
       WSREP_INFO("Slave error due to node going non-primary");
       WSREP_INFO("wsrep_restart_slave was set and therefore slave will be "
                  "automatically restarted when node joins back to cluster");
-      wsrep_restart_slave_activated= TRUE;
-    }
+      if (wsrep_wait_ready(thd))
+      {
+        wsrep_close(thd);
+        delete serial_rgi;
+        if (thd_initialized)
+          server_threads.erase(thd);
+        delete thd;
+        goto wsrep_restart_point;
+      }
+      else
+      {
+        /* The node is being shutdown. Fallthrough. */
+        mysql_mutex_lock(&rli->run_lock);
+      }
   }
   wsrep_close(thd);
 #endif /* WITH_WSREP */