First implementation of instance-manager carcass:

logging, options, angel process.


configure.in:
  added support for instance-manager directory
BitKeeper/etc/logging_ok:
  Logging to logging@openlogging.org accepted

1 files changed, 249 insertions, 0 deletions

diff --git a/server-tools/instance-manager/mysqlmanager.cc b/server-tools/instance-manager/mysqlmanager.cc
new file mode 100644
index 00000000000..a08b0691eec
--- /dev/null
+++ b/server-tools/instance-manager/mysqlmanager.cc
@@ -0,0 +1,249 @@
+/* Copyright (C) 2000 MySQL AB & MySQL Finland AB & TCX DataKonsult AB
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */
+
+#include <my_global.h>
+#include <my_sys.h>
+#include <string.h>
+#include <signal.h>
+#include <sys/wait.h>
+
+#include "manager.h"
+#include "options.h"
+#include "log.h"
+
+/*
+  Few notes about the Instance Manager architecture:
+  Instance Manager consisits of two processes: the angel process, and the
+  instance manager process. Responsibilities of the angel process is to
+  monitor the instance manager process, and restart it in case of
+  failure/shutdown. The angel process is started only if startup option
+  '--run-as-service' is provided.
+  The Instance Manager process consists of several
+  subsystems (thread sets):
+  - the signal handling thread: it's responsibilities are to handle 
+    user signals and propogate them to the other threads. All other threads
+    are accounted in the signal handler thread Thread Repository.
+  - the listener: listens all sockets. There is a listening 
+    socket for each (mysql, http, snmp, rendezvous (?)) subsystem.
+  - mysql subsystem: Instance Manager acts like an ordinary MySQL Server,
+    but with very restricted command set. Each MySQL client connection is 
+    handled in a separate thread. All MySQL client connections threads
+    constitute mysql subsystem.
+  - http subsystem: it is also possible to talk with Instance Manager via 
+    http. One thread per http connection is used. Threads are pooled.
+  - 'snmp' connections (FIXME: I know nothing about it yet)
+  - rendezvous threads 
+*/
+
+static void init_environment(char *progname);
+static void daemonize(const char *log_file_name);
+static void angel(const Options &options);
+
+
+/*
+  main, entry point
+  - init environment
+  - handle options
+  - daemonize and run angel process (if necessary)
+  - run manager process
+*/
+
+int main(int argc, char *argv[])
+{
+  init_environment(argv[0]);
+  Options options;
+  options.load(argc, argv);
+  if (options.run_as_service)
+  {
+    daemonize(options.log_file_name);
+    angel(options);
+  }
+  else
+    manager(options.log_file_name);
+  return 0;
+}
+
+/******************* Auxilary functions implementation **********************/
+
+
+/*
+  Init environment, common for daemon and non-daemon
+*/
+
+static void init_environment(char *progname)
+{
+  MY_INIT(progname);
+  log_init();
+  umask(0117);
+}
+
+
+/*
+  Become a UNIX service
+  SYNOPSYS
+    daemonize()
+*/
+
+static void daemonize(const char *log_file_name)
+{
+  pid_t pid= fork();
+  switch (pid) {
+  case -1:                                      // parent, fork error
+    die("daemonize(): fork failed, %s", strerror(errno));
+  case 0:                                       // child, fork ok
+    int fd;
+    /*
+      Become a session leader: setsid should succeed because child is
+      guaranteed not to be a process group leader (it belongs to the 
+      process group of the parent.) 
+      The goal is not to have a controlling terminal.
+    */
+    if (setsid() == -1)
+      die("daemonize(): setsid() failed, %s", strerror(errno)); 
+    /*
+      As we now don't have a controlling terminal we will not receive
+      tty-related signals - no need to ignore them.
+    */
+
+    close(STDIN_FILENO);
+
+    fd= open(log_file_name, O_WRONLY | O_CREAT | O_APPEND | O_NOCTTY, 
+                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
+    if (fd < 0)
+      die("daemonize(): failed to open log file %s, %s", log_file_name,
+          strerror(errno));
+    dup2(fd, STDOUT_FILENO);
+    dup2(fd, STDERR_FILENO);
+    if (fd != STDOUT_FILENO && fd != STDERR_FILENO)
+      close(fd);
+
+    /* TODO: chroot() and/or chdir() here */
+    break;
+  default:                                      
+    /* successfully exit from parent */
+    exit(0);
+  }
+}
+
+
+enum { CHILD_OK= 0, CHILD_NEED_RESPAWN, CHILD_EXIT_ANGEL };
+
+static volatile sig_atomic_t child_status= CHILD_OK;
+
+/* 
+  Signal handler for SIGCHLD: reap child, analyze child exit status, and set
+  child_status appropriately.
+*/
+
+void reap_child(int __attribute__((unused)) signo)
+{ 
+  int child_exit_status;
+  /* As we have only one child, no need to cycle waitpid */
+  if (waitpid(0, &child_exit_status, WNOHANG) > 0)
+  {
+    if (WIFSIGNALED(child_exit_status))
+      child_status= CHILD_NEED_RESPAWN;
+    else
+      /*
+        As we reap_child is not called for SIGSTOP, we should be here only
+        if the child exited normally.
+      */
+      child_status= CHILD_EXIT_ANGEL;
+  }
+}
+
+/* Not static to reuse it in childs */
+
+volatile sig_atomic_t is_terminated= 0;
+
+/*
+  Signal handler for terminate signals - SIGTERM, SIGHUP, SIGINT.
+  Set termination status and return.
+  (q) do we need to handle SIGQUIT?
+*/
+
+void terminate(int signo)
+{
+  is_terminated= signo; 
+}
+
+
+/*
+  Fork a child and monitor it.
+  User can explicitly kill the angel process with SIGTERM/SIGHUP/SIGINT.
+  Angel process will exit silently if mysqlmanager exits normally.
+*/
+
+static void angel(const Options &options)
+{
+  /* install signal handlers */
+  sigset_t zeromask;                            // to sigsuspend in parent
+  struct sigaction sa_chld, sa_term;
+  struct sigaction sa_chld_out, sa_term_out, sa_int_out, sa_hup_out;
+  if (sigemptyset(&zeromask) ||
+      sigemptyset(&sa_chld.sa_mask) ||
+      sigemptyset(&sa_term.sa_mask))            // how can it fail?
+    die("angel(): sigemptyset() failed, %s", strerror(errno));
+  sa_chld.sa_handler= reap_child;
+  sa_chld.sa_flags= SA_NOCLDSTOP;
+  sa_term.sa_handler= terminate;
+  sa_term.sa_flags= 0;
+  if (sigaction(SIGCHLD, &sa_chld, &sa_chld_out) == -1 ||
+      sigaction(SIGTERM, &sa_term, &sa_term_out) == -1 ||
+      sigaction(SIGINT, &sa_term, &sa_int_out) == -1 ||
+      sigaction(SIGHUP, &sa_term, &sa_hup_out) == -1)
+    die("angel(): sigaction() failed, %s", strerror(errno));
+
+  /* spawn a child */
+spawn:
+  pid_t pid= fork();
+  switch (pid) {
+  case -1:
+    die("angel(): fork failed, %s", strerror(errno)); 
+  case 0:                                     // child, success
+    /*
+      restore default actions for signals to let the manager work with
+      signals as he wishes
+    */ 
+    if (sigaction(SIGCHLD, &sa_chld_out, 0) == -1 ||
+        sigaction(SIGTERM, &sa_term_out, 0) == -1 ||
+        sigaction(SIGINT, &sa_int_out, 0) == -1 ||
+        sigaction(SIGHUP, &sa_hup_out, 0) == -1)
+      die("angel(): child failed to restore old signal actions, %s",
+          strerror(errno));
+
+    manager(options.socket_file_name);
+  default:                                    // parent, success
+    while (child_status == CHILD_OK && is_terminated == 0)
+    {
+      errno= 0;
+      sigsuspend(&zeromask);
+      if (errno != EINTR)
+        die("angel(): sigsuspend failed, %s", strerror(errno));
+    }
+    if (is_terminated)
+      log_info("angel got signal %d (%s), exiting",
+               is_terminated, sys_siglist[is_terminated]);
+    else if (child_status == CHILD_NEED_RESPAWN)
+    { 
+      child_status= CHILD_OK;
+      log_error("angel(): mysqlmanager exited abnormally: respawning...");
+      sleep(1); /* don't respawn too fast */
+      goto spawn;
+    }
+    /* mysqlmanager successfully exited, let's silently evaporate */
+  }
+}