database/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-08-01 03:47:19 +03:00
Code Activity

more fixes for IM to substitude mysqld_safe in startup scripts

Browse Source 
BitKeeper/deleted/.del-thread_repository.cc~bba09f64f8cb4037:
  Delete: server-tools/instance-manager/thread_repository.cc
BitKeeper/deleted/.del-thread_repository.h~e6a3b9cab7a6612a:
  Delete: server-tools/instance-manager/thread_repository.h
server-tools/instance-manager/commands.cc:
  All instances are guarded by default now, so we need to perform check on whether the instance is nonguarded,
  rather then guarded when adding it to the list of guarded instnces.
server-tools/instance-manager/guardian.cc:
  Guardian rewritten to start instances by default, and shut them down, when exiting. Behaviour of the guardian
  in case of the instance crash has changed. Now it tries to restart an instance constantly in the first 2 
  seconds after the crash was noticed, and then it tries restart an instance once in the MONITORING_INTERVAL.
  If it failed to restart instance for "restart_retry" (compiled-in value) times, guardian stops trying to
  restart it.
server-tools/instance-manager/guardian.h:
  Several new functions and variables declared.
server-tools/instance-manager/instance.cc:
  now start doesn't call stop(), but rather tries to remove the pidfile itself
server-tools/instance-manager/instance.h:
  cleanup
server-tools/instance-manager/instance_map.cc:
  no more "admin" options
server-tools/instance-manager/instance_map.h:
  User and password purged from instance_map options, as IM shouldn't know them
server-tools/instance-manager/instance_options.cc:
  new option added -- shutdown_delay, guarded option now called nonguaded and has the opposite meaning
server-tools/instance-manager/instance_options.h:
  appropriate changes, reflecting options addition/removal
server-tools/instance-manager/manager.cc:
  shutdown process is complicated a bit -- at first signal thread should stop guardian, and only then the IM
  itself
server-tools/instance-manager/messages.cc:
  update error message
server-tools/instance-manager/options.cc:
  admin user/password purged from mysqlmanager options
This commit is contained in:
unknown
2005-01-31 23:54:08 +03:00
parent 28b1aefa00
commit 92a52cccf6
14 changed files with 160 additions and 369 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
server-tools/instance-manager/instance.cc
View File

@ -42,10 +42,16 @@
int Instance::start()
{
/* echk for the pidfile and remove it */
pid_t pid;
if (!is_running())
{
stop();
if ((pid= options.get_pid()) != 0) /* check the pidfile */
if (options.unlink_pidfile()) /* remove stalled pidfile */
log_error("cannot remove pidfile for instance %i, this might be \
since IM lacks permmissions or hasn't found the pidifle",
options.instance_name);
log_info("starting instance %s", options.instance_name);
switch (pid= fork()) {
case 0:
@ -145,14 +151,17 @@ int Instance::stop()
{
pid_t pid;
struct timespec timeout;
time_t waitchild= 35; /* */
int waitchild= DEFAULT_SHUTDOWN_DELAY;
if (options.shutdown_delay != NULL)
waitchild= atoi(options.shutdown_delay);
if ((pid= options.get_pid()) != 0) /* get pid from pidfile */
{
/*
If we cannot kill mysqld, then it has propably crashed.
Let us try to remove staled pidfile and return succes as mysqld
is stopped
is probably stopped
*/
if (kill(pid, SIGTERM))
{
@ -161,9 +170,6 @@ int Instance::stop()
since IM lacks permmissions or hasn't found the pidifle",
options.instance_name);
log_error("The instance %s has probably crashed or IM lacks permissions \
to kill it. in either case something seems to be wrong. \
Check your setup", options.instance_name);
return 0;
}
@ -174,7 +180,7 @@ int Instance::stop()
if (pthread_mutex_lock(&instance_map->pid_cond.LOCK_pid))
goto err; /* perhaps this should be procecced differently */
while (options.get_pid() != 0)
while (options.get_pid() != 0) /* while server isn't stopped */
{
int status;