[pgpool-hackers: 378] Re: Exception Manager for pgpool
Tatsuo Ishii
ishii at postgresql.org
Wed Oct 16 09:22:15 JST 2013
> Hi Tatsuo
>
> I have been trying to reproduce this compilation issue but unable to do so.
> I have compiled the patched code with GCC 4.6.3 and also with
> LLVM 5.0 and in both case I am able to successfully compile the code.
> Can you please guide me to reproduce the issue at my end as the Oid
> definition which is conflicting at your end is not changed by the patch.
No idea. I just type "make" and got the error.
> here is the gcc log of files compilation,
>
> gcc -DHAVE_CONFIG_H
> -DDEFAULT_CONFIGDIR=\"/home/usama/EDB/pgpool/elog_work/mypgpool/test/comanddel/installed/etc\"
> -I. -I../src/include -D_GNU_SOURCE -I /home/usama/EDB/PG/installed/include
> -O0 -g3 -Wall -Wmissing-prototypes -Wmissing-declarations -MT main/main.o
> -MD -MP -MF $depbase.Tpo -c -o main/main.o main/main.c &&\
> mv -f $depbase.Tpo $depbase.Po
> main/main.c: In function ‘daemonize’:
> main/main.c:384:17: warning: variable ‘rc_chdir’ set but not used
> [-Wunused-but-set-variable]
> depbase=`echo main/pgpool_main.o | sed 's|[^/]*$|.deps/&|;s|\.o$||'`;\
> gcc -DHAVE_CONFIG_H
> -DDEFAULT_CONFIGDIR=\"/home/usama/EDB/pgpool/elog_work/mypgpool/test/comanddel/installed/etc\"
> -I. -I../src/include -D_GNU_SOURCE -I /home/usama/EDB/PG/installed/include
> -O0 -g3 -Wall -Wmissing-prototypes -Wmissing-declarations -MT
> main/pgpool_main.o -MD -MP -MF $depbase.Tpo -c -o main/pgpool_main.o
> main/pgpool_main.c &&\
> mv -f $depbase.Tpo $depbase.Po
> main/pgpool_main.c: In function ‘find_primary_node’:
> main/pgpool_main.c:1981:14: warning: variable ‘status’ set but not used
> [-Wunused-but-set-variable]
> ..
>
> pgpool2$ gcc --version
> gcc (Ubuntu/Linaro 4.6.3-1ubuntu5) 4.6.3
> Copyright (C) 2011 Free Software Foundation, Inc.
>
> And on OS/X the compilation is also successfull.
> gcc --version
> Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr
> --with-gxx-include-dir=/usr/include/c++/4.2.1
> Apple LLVM version 5.0 (clang-500.2.76) (based on LLVM 3.3svn)
> Target: x86_64-apple-darwin12.5.0
>
> Thanks
> Usama
>
>
>
> On Sat, Oct 12, 2013 at 5:41 AM, Tatsuo Ishii <ishii at postgresql.org> wrote:
>
>> I have freshly checkout git master and applied your patch.
>> I got this:
>>
>> gcc -DHAVE_CONFIG_H -DDEFAULT_CONFIGDIR=\"/home/t-ishii/work/
>> git.postgresql.org/exception/pgpool2/src/test/regression/temp/installed/etc\"
>> -I. -I../src/include -D_GNU_SOURCE -I /usr/local/pgsql/include -g -O2
>> -Wall -Wmissing-prototypes -Wmissing-declarations -MT main/pgpool_main.o
>> -MD -MP -MF $depbase.Tpo -c -o main/pgpool_main.o main/pgpool_main.c &&\
>> mv -f $depbase.Tpo $depbase.Po
>> In file included from /usr/local/pgsql/include/libpq-fe.h:29,
>> from ../src/include/pool.h:34,
>> from main/pgpool_main.c:51:
>> /usr/local/pgsql/include/postgres_ext.h:31: error: redefinition of typedef
>> ‘PoolOid’
>> ../src/include/parser/pool_parser.h:43: note: previous declaration of
>> ‘PoolOid’ was here
>>
>> This is caused by the conflict in postgres_ext.h:
>> typedef unsigned int Oid;
>>
>> and
>>
>> ./include/parser/pool_parser.h:typedef unsigned int PoolOid;
>> ./include/parser/pool_parser.h:#define Oid PoolOid
>>
>> This is gcc 4.4.5.
>> --
>> Tatsuo Ishii
>> SRA OSS, Inc. Japan
>> English: http://www.sraoss.co.jp/index_en.php
>> Japanese: http://www.sraoss.co.jp
>>
>> > Hi
>> >
>> > Seems like the repository was changed after I had generated the patch.
>> > Please find the updated patch rebased with the current state of git.
>> >
>> >
>> > Thanks
>> > Usama
>> >
>> >
>> >
>> > On Fri, Oct 11, 2013 at 9:47 AM, Ahsan Hadi <ahsan.hadi at enterprisedb.com
>> >wrote:
>> >
>> >> Usama,
>> >> Please take a look.
>> >>
>> >> Tatsuo,
>> >> I was apply to apply this patch cleanly on pgpool master and test before
>> >> Usama sent the patch. It is likely that we have done some changes to the
>> >> master branch since then..
>> >>
>> >>
>> >> On Fri, Oct 11, 2013 at 4:03 AM, Tatsuo Ishii <ishii at postgresql.org
>> >wrote:
>> >>
>> >>> Usama,
>> >>>
>> >>> I have applied your patch and got following error:
>> >>>
>> >>> Hunk #12 FAILED at 499.
>> >>> 1 out of 12 hunks FAILED -- saving rejects to file src/main/main.c.rej
>> >>>
>> >>> I also attached main.c.rej.
>> >>> --
>> >>> Tatsuo Ishii
>> >>> SRA OSS, Inc. Japan
>> >>> English: http://www.sraoss.co.jp/index_en.php
>> >>> Japanese: http://www.sraoss.co.jp
>> >>>
>> >>> > Muhammad,
>> >>> >
>> >>> > Thank you for your great work! I'll look into this.
>> >>> > --
>> >>> > Tatsuo Ishii
>> >>> > SRA OSS, Inc. Japan
>> >>> > English: http://www.sraoss.co.jp/index_en.php
>> >>> > Japanese: http://www.sraoss.co.jp
>> >>> >
>> >>> >> Hi
>> >>> >>
>> >>> >> I am working on adding the exception manager in pgpool and my plan
>> of
>> >>> >> action for this is to use the Postgres exception manager API (elog
>> and
>> >>> >> friends). Since the exception manager in Postgres uses the long
>> jump,
>> >>> so
>> >>> >> importing this API in pgpool will effect all existing pgpool code
>> flows
>> >>> >> especially in case of an error. and lot of care will be required for
>> >>> this
>> >>> >> integration, and secondly the elog API along with its friends will
>> >>> touch
>> >>> >> almost all parts of pgpool source code which will add up to a very
>> huge
>> >>> >> patch.
>> >>> >> So instead throwing a very huge patch to the community my plan is to
>> >>> divide
>> >>> >> this task into multiple smaller sub tasks so that it would be easier
>> >>> >> maintain and review the patch.
>> >>> >>
>> >>> >> Cut to the chase, attached is the first of the series of related
>> >>> patches
>> >>> >> to come.This is the first cut patch for implementing the exception
>> >>> manager
>> >>> >> in pgpool. As described above the exception manager and related
>> code is
>> >>> >> borrowed from PostgreSQL source code.
>> >>> >> and the exception manager (elog API) is very closely tied with
>> memory
>> >>> >> manager in PostgreSQL (palloc API) so the patch also borrows the
>> PG's
>> >>> >> memory manager.
>> >>> >>
>> >>> >> Below is the little description of things part of this patch.
>> >>> >>
>> >>> >> -- Exception manager API of Postgres is added to pgpool, The API
>> >>> consists
>> >>> >> of elog.c and elog.h files. Since this API is very extensive and is
>> >>> >> designed for PostgreSQL so to fit it properly into pgpool I have
>> >>> modified
>> >>> >> it a little bit, and most of the modifications are related to
>> removal
>> >>> of
>> >>> >> code which is not required for pgpool.
>> >>> >>
>> >>> >> -- Added on_proc_exit callback mechanism of Postgres. To facilitate
>> the
>> >>> >> cleanup at exit time.
>> >>> >>
>> >>> >> -- Added PostgreSQL's memory manager (palloc API). This includes the
>> >>> client
>> >>> >> side palloc functions placed in 'src/tools' directory (fe_memutils)
>> >>> >>
>> >>> >> -- Removed the existing memory manager which was very minimalistic
>> and
>> >>> was
>> >>> >> not integrated in all parts of the code.
>> >>> >>
>> >>> >> -- I have also tried to reflector some portions of code to make the
>> >>> code
>> >>> >> more readable at first glance. This includes
>> >>> >>
>> >>> >> - dividing the main.c file into two files main.c and pgpool_main.c,
>> >>> Now the
>> >>> >> main.c file only contains the code related to early initialisations
>> of
>> >>> >> pgpool and parsing the command line options and related code. The
>> >>> actual
>> >>> >> logic of the pgpool main process is moved to new pgpool_main.c file.
>> >>> >> - breaking up some large functions in child.c into smaller
>> functions.
>> >>> >> - rewrite the pgpool's main loop logic to make the code more
>> readable.
>> >>> >>
>> >>> >>
>> >>> >> Remaining TODOs on this front.
>> >>> >>
>> >>> >> -- The current patch only integrates the memory and exception
>> manager
>> >>> in
>> >>> >> main process and connection creation segment of pgpool child
>> process.
>> >>> >> integration of newly added APIs in pcp and worker child process
>> codes
>> >>> will
>> >>> >> be done be next patch.
>> >>> >>
>> >>> >> -- Integration of newly added API into query processor logic in
>> child
>> >>> >> process. ( this will be the toughest part)
>> >>> >>
>> >>> >> -- elog.c and elog.h files needs some cleanups and changes ( to
>> remove
>> >>> >> unwanted functions and data members of ErrorData structure) but this
>> >>> will
>> >>> >> be done at the end when we will have 100% surety if something in
>> there
>> >>> is
>> >>> >> required or not.
>> >>> >>
>> >>> >>
>> >>> >> Thanks
>> >>> >> Muhammad Usama
>> >>> > _______________________________________________
>> >>> > pgpool-hackers mailing list
>> >>> > pgpool-hackers at pgpool.net
>> >>> > http://www.pgpool.net/mailman/listinfo/pgpool-hackers
>> >>>
>> >>> --- src/main/main.c
>> >>> +++ src/main/main.c
>> >>> @@ -499,1915 +-23,60 @@
>> >>> fd = open(pool_config->pid_file_name, O_CREAT|O_WRONLY,
>> >>> S_IRUSR|S_IWUSR);
>> >>> if (fd == -1)
>> >>> {
>> >>> - pool_error("could not open pid file as %s. reason: %s",
>> >>> - pool_config->pid_file_name,
>> >>> strerror(errno));
>> >>> - pool_shmem_exit(1);
>> >>> - exit(1);
>> >>> + ereport(FATAL,
>> >>> + (errmsg("could not open pid file as %s. reason:
>> >>> %s",
>> >>> + pool_config->pid_file_name,
>> >>> strerror(errno))));
>> >>> }
>> >>> snprintf(pidbuf, sizeof(pidbuf), "%d", (int)getpid());
>> >>> if (write(fd, pidbuf, strlen(pidbuf)+1) == -1)
>> >>> {
>> >>> - pool_error("could not write pid file as %s. reason:
>> %s",
>> >>> - pool_config->pid_file_name,
>> >>> strerror(errno));
>> >>> close(fd);
>> >>> - pool_shmem_exit(1);
>> >>> - exit(1);
>> >>> + ereport(FATAL,
>> >>> + (errmsg("could not write pid file as %s.
>> reason:
>> >>> %s",
>> >>> + pool_config->pid_file_name,
>> >>> strerror(errno))));
>> >>> }
>> >>> if (fsync(fd) == -1)
>> >>> {
>> >>> - pool_error("could not fsync pid file as %s. reason:
>> %s",
>> >>> - pool_config->pid_file_name,
>> >>> strerror(errno));
>> >>> close(fd);
>> >>> - pool_shmem_exit(1);
>> >>> - exit(1);
>> >>> + ereport(FATAL,
>> >>> + (errmsg("could not fsync pid file as %s.
>> reason:
>> >>> %s",
>> >>> + pool_config->pid_file_name,
>> >>> strerror(errno))));
>> >>> }
>> >>> if (close(fd) == -1)
>> >>> {
>> >>> - pool_error("could not close pid file as %s. reason:
>> %s",
>> >>> - pool_config->pid_file_name,
>> >>> strerror(errno));
>> >>> - pool_shmem_exit(1);
>> >>> - exit(1);
>> >>> + ereport(FATAL,
>> >>> + (errmsg("could not close pid file as %s.
>> reason:
>> >>> %s",
>> >>> + pool_config->pid_file_name,
>> >>> strerror(errno))));
>> >>> }
>> >>> + /* register the call back to delete the pid file at system
>> exit */
>> >>> + on_proc_exit(FileUnlink, (Datum) pool_config->pid_file_name);
>> >>> }
>> >>>
>> >>> /*
>> >>> -* Read the status file
>> >>> -*/
>> >>> -static int read_status_file(bool discard_status)
>> >>> + * get_config_file_name: return full path of pgpool.conf.
>> >>> + */
>> >>> +char *get_config_file_name(void)
>> >>> {
>> >>> - FILE *fd;
>> >>> - char fnamebuf[POOLMAXPATHLEN];
>> >>> - int i;
>> >>> - bool someone_wakeup = false;
>> >>> -
>> >>> - snprintf(fnamebuf, sizeof(fnamebuf), "%s/%s",
>> >>> pool_config->logdir, STATUS_FILE_NAME);
>> >>> - fd = fopen(fnamebuf, "r");
>> >>> - if (!fd)
>> >>> - {
>> >>> - pool_log("Backend status file %s does not exist",
>> >>> fnamebuf);
>> >>> - return -1;
>> >>> - }
>> >>> -
>> >>> - /*
>> >>> - * If discard_status is true, unlink pgpool_status and
>> >>> - * do not restore previous status.
>> >>> - */
>> >>> - if (discard_status)
>> >>> - {
>> >>> - fclose(fd);
>> >>> - if (unlink(fnamebuf) == 0)
>> >>> - {
>> >>> - pool_log("Backend status file %s discarded",
>> >>> fnamebuf);
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - pool_error("Failed to discard backend status
>> file
>> >>> %s reason:%s", fnamebuf, strerror(errno));
>> >>> - }
>> >>> - return 0;
>> >>> - }
>> >>> -
>> >>> - if (fread(&backend_rec, 1, sizeof(backend_rec), fd) !=
>> >>> sizeof(backend_rec))
>> >>> - {
>> >>> - pool_error("Could not read backend status file as %s.
>> >>> reason: %s",
>> >>> - fnamebuf, strerror(errno));
>> >>> - fclose(fd);
>> >>> - return -1;
>> >>> - }
>> >>> - fclose(fd);
>> >>> -
>> >>> - for (i=0;i< pool_config->backend_desc->num_backends;i++)
>> >>> - {
>> >>> - if (backend_rec.status[i] == CON_DOWN)
>> >>> - {
>> >>> - BACKEND_INFO(i).backend_status = CON_DOWN;
>> >>> - pool_log("read_status_file: %d th backend is
>> set
>> >>> to down status", i);
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - BACKEND_INFO(i).backend_status =
>> CON_CONNECT_WAIT;
>> >>> - someone_wakeup = true;
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - /*
>> >>> - * If no one woke up, we regard the status file bogus
>> >>> - */
>> >>> - if (someone_wakeup == false)
>> >>> - {
>> >>> - for (i=0;i<
>> pool_config->backend_desc->num_backends;i++)
>> >>> - {
>> >>> - BACKEND_INFO(i).backend_status =
>> CON_CONNECT_WAIT;
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - return 0;
>> >>> + return conf_file;
>> >>> }
>> >>>
>> >>> /*
>> >>> -* Write the pid file
>> >>> -*/
>> >>> -static int write_status_file(void)
>> >>> -{
>> >>> - FILE *fd;
>> >>> - char fnamebuf[POOLMAXPATHLEN];
>> >>> - int i;
>> >>> -
>> >>> - snprintf(fnamebuf, sizeof(fnamebuf), "%s/%s",
>> >>> pool_config->logdir, STATUS_FILE_NAME);
>> >>> - fd = fopen(fnamebuf, "w");
>> >>> - if (!fd)
>> >>> - {
>> >>> - pool_error("Could not open status file %s", fnamebuf);
>> >>> - return -1;
>> >>> - }
>> >>> -
>> >>> - memset(&backend_rec, 0, sizeof(backend_rec));
>> >>> -
>> >>> - for (i=0;i< pool_config->backend_desc->num_backends;i++)
>> >>> - {
>> >>> - backend_rec.status[i] = BACKEND_INFO(i).backend_status;
>> >>> - }
>> >>> -
>> >>> - if (fwrite(&backend_rec, 1, sizeof(backend_rec), fd) !=
>> >>> sizeof(backend_rec))
>> >>> - {
>> >>> - pool_error("Could not write backend status file as %s.
>> >>> reason: %s",
>> >>> - fnamebuf, strerror(errno));
>> >>> - fclose(fd);
>> >>> - return -1;
>> >>> - }
>> >>> - fclose(fd);
>> >>> - return 0;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * fork a child for PCP
>> >>> - */
>> >>> -pid_t pcp_fork_a_child(int unix_fd, int inet_fd, char *pcp_conf_file)
>> >>> -{
>> >>> - pid_t pid;
>> >>> -
>> >>> - pid = fork();
>> >>> -
>> >>> - if (pid == 0)
>> >>> - {
>> >>> - close(pipe_fds[0]);
>> >>> - close(pipe_fds[1]);
>> >>> -
>> >>> - myargv = save_ps_display_args(myargc, myargv);
>> >>> -
>> >>> - /* call PCP child main */
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> - health_check_timer_expired = 0;
>> >>> - reload_config_request = 0;
>> >>> - run_as_pcp_child = true;
>> >>> - pcp_do_child(unix_fd, inet_fd, pcp_conf_file);
>> >>> - }
>> >>> - else if (pid == -1)
>> >>> - {
>> >>> - pool_error("fork() failed. reason: %s",
>> strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> - return pid;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> -* fork a child
>> >>> -*/
>> >>> -pid_t fork_a_child(int unix_fd, int inet_fd, int id)
>> >>> -{
>> >>> - pid_t pid;
>> >>> -
>> >>> - pid = fork();
>> >>> -
>> >>> - if (pid == 0)
>> >>> - {
>> >>> - /* Before we unconditionally closed pipe_fds[0] and
>> >>> pipe_fds[1]
>> >>> - * here, which is apparently wrong since in the start
>> up
>> >>> of
>> >>> - * pgpool, pipe(2) is not called yet and it mistakenly
>> >>> closes
>> >>> - * fd 0. Now we check the fd > 0 before close(),
>> expecting
>> >>> - * pipe returns fds greater than 0. Note that we
>> cannot
>> >>> - * unconditionally remove close(2) calls since
>> >>> fork_a_child()
>> >>> - * may be called *after* pgpool starting up.
>> >>> - */
>> >>> - if (pipe_fds[0] > 0)
>> >>> - {
>> >>> - close(pipe_fds[0]);
>> >>> - close(pipe_fds[1]);
>> >>> - }
>> >>> -
>> >>> - myargv = save_ps_display_args(myargc, myargv);
>> >>> -
>> >>> - /* call child main */
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> - health_check_timer_expired = 0;
>> >>> - reload_config_request = 0;
>> >>> - my_proc_id = id;
>> >>> - run_as_pcp_child = false;
>> >>> - do_child(unix_fd, inet_fd);
>> >>> - }
>> >>> - else if (pid == -1)
>> >>> - {
>> >>> - pool_error("fork() failed. reason: %s",
>> strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> - return pid;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> -* fork worker child process
>> >>> -*/
>> >>> -pid_t worker_fork_a_child()
>> >>> -{
>> >>> - pid_t pid;
>> >>> -
>> >>> - pid = fork();
>> >>> -
>> >>> - if (pid == 0)
>> >>> - {
>> >>> - /* Before we unconditionally closed pipe_fds[0] and
>> >>> pipe_fds[1]
>> >>> - * here, which is apparently wrong since in the start
>> up
>> >>> of
>> >>> - * pgpool, pipe(2) is not called yet and it mistakenly
>> >>> closes
>> >>> - * fd 0. Now we check the fd > 0 before close(),
>> expecting
>> >>> - * pipe returns fds greater than 0. Note that we
>> cannot
>> >>> - * unconditionally remove close(2) calls since
>> >>> fork_a_child()
>> >>> - * may be called *after* pgpool starting up.
>> >>> - */
>> >>> - if (pipe_fds[0] > 0)
>> >>> - {
>> >>> - close(pipe_fds[0]);
>> >>> - close(pipe_fds[1]);
>> >>> - }
>> >>> -
>> >>> - myargv = save_ps_display_args(myargc, myargv);
>> >>> -
>> >>> - /* call child main */
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> - health_check_timer_expired = 0;
>> >>> - reload_config_request = 0;
>> >>> - do_worker_child();
>> >>> - }
>> >>> - else if (pid == -1)
>> >>> - {
>> >>> - pool_error("fork() failed. reason: %s",
>> strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> - return pid;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> -* create inet domain socket
>> >>> -*/
>> >>> -static int create_inet_domain_socket(const char *hostname, const int
>> >>> port)
>> >>> -{
>> >>> - struct sockaddr_in addr;
>> >>> - int fd;
>> >>> - int status;
>> >>> - int one = 1;
>> >>> - int len;
>> >>> - int backlog;
>> >>> -
>> >>> - fd = socket(AF_INET, SOCK_STREAM, 0);
>> >>> - if (fd == -1)
>> >>> - {
>> >>> - pool_error("Failed to create INET domain socket.
>> reason:
>> >>> %s", strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> - if ((setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (char *) &one,
>> >>> - sizeof(one))) == -1)
>> >>> - {
>> >>> - pool_error("setsockopt() failed. reason: %s",
>> >>> strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> -
>> >>> - memset((char *) &addr, 0, sizeof(addr));
>> >>> - addr.sin_family = AF_INET;
>> >>> -
>> >>> - if (strcmp(hostname, "*")==0)
>> >>> - {
>> >>> - addr.sin_addr.s_addr = htonl(INADDR_ANY);
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - struct hostent *hostinfo;
>> >>> -
>> >>> - hostinfo = gethostbyname(hostname);
>> >>> - if (!hostinfo)
>> >>> - {
>> >>> - pool_error("could not resolve host name \"%s\":
>> >>> %s", hostname, hstrerror(h_errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> - addr.sin_addr = *(struct in_addr *) hostinfo->h_addr;
>> >>> - }
>> >>> -
>> >>> - addr.sin_port = htons(port);
>> >>> - len = sizeof(struct sockaddr_in);
>> >>> - status = bind(fd, (struct sockaddr *)&addr, len);
>> >>> - if (status == -1)
>> >>> - {
>> >>> - char *host = "", *serv = "";
>> >>> - char hostname[NI_MAXHOST], servname[NI_MAXSERV];
>> >>> - if (getnameinfo((struct sockaddr *) &addr, len,
>> hostname,
>> >>> sizeof(hostname), servname, sizeof(servname), 0) == 0) {
>> >>> - host = hostname;
>> >>> - serv = servname;
>> >>> - }
>> >>> - pool_error("bind(%s:%s) failed. reason: %s", host,
>> serv,
>> >>> strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> -
>> >>> - backlog = pool_config->num_init_children * 2;
>> >>> - if (backlog > PGPOOLMAXLITSENQUEUELENGTH)
>> >>> - backlog = PGPOOLMAXLITSENQUEUELENGTH;
>> >>> -
>> >>> - status = listen(fd, backlog);
>> >>> - if (status < 0)
>> >>> - {
>> >>> - pool_error("listen() failed. reason: %s",
>> >>> strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> - return fd;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> -* create UNIX domain socket
>> >>> -*/
>> >>> -static int create_unix_domain_socket(struct sockaddr_un un_addr_tmp)
>> >>> -{
>> >>> - struct sockaddr_un addr;
>> >>> - int fd;
>> >>> - int status;
>> >>> - int len;
>> >>> -
>> >>> - fd = socket(AF_UNIX, SOCK_STREAM, 0);
>> >>> - if (fd == -1)
>> >>> - {
>> >>> - pool_error("Failed to create UNIX domain socket.
>> reason:
>> >>> %s", strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> - memset((char *) &addr, 0, sizeof(addr));
>> >>> - addr.sun_family = AF_UNIX;
>> >>> - snprintf(addr.sun_path, sizeof(addr.sun_path), "%s",
>> >>> un_addr_tmp.sun_path);
>> >>> - len = sizeof(struct sockaddr_un);
>> >>> - status = bind(fd, (struct sockaddr *)&addr, len);
>> >>> - if (status == -1)
>> >>> - {
>> >>> - pool_error("bind(%s) failed. reason: %s",
>> addr.sun_path,
>> >>> strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> -
>> >>> - if (chmod(un_addr_tmp.sun_path, 0777) == -1)
>> >>> - {
>> >>> - pool_error("chmod() failed. reason: %s",
>> strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> -
>> >>> - status = listen(fd, PGPOOLMAXLITSENQUEUELENGTH);
>> >>> - if (status < 0)
>> >>> - {
>> >>> - pool_error("listen() failed. reason: %s",
>> >>> strerror(errno));
>> >>> - myexit(1);
>> >>> - }
>> >>> - return fd;
>> >>> -}
>> >>> -
>> >>> -static void myunlink(const char* path)
>> >>> -{
>> >>> - if (unlink(path) == 0) return;
>> >>> - pool_error("unlink(%s) failed: %s", path, strerror(errno));
>> >>> -}
>> >>> -
>> >>> -static void myexit(int code)
>> >>> -{
>> >>> - int i;
>> >>> -
>> >>> - if (getpid() != mypid)
>> >>> - return;
>> >>> -
>> >>> - if (process_info != NULL) {
>> >>> - POOL_SETMASK(&AuthBlockSig);
>> >>> - exiting = 1;
>> >>> - for (i = 0; i < pool_config->num_init_children; i++)
>> >>> - {
>> >>> - pid_t pid = process_info[i].pid;
>> >>> - if (pid)
>> >>> - {
>> >>> - kill(pid, SIGTERM);
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - /* wait for all children to exit */
>> >>> - while (wait(NULL) > 0)
>> >>> - ;
>> >>> - if (errno != ECHILD)
>> >>> - pool_error("wait() failed. reason:%s",
>> >>> strerror(errno));
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> - }
>> >>> -
>> >>> - myunlink(un_addr.sun_path);
>> >>> - myunlink(pcp_un_addr.sun_path);
>> >>> - myunlink(pool_config->pid_file_name);
>> >>> -
>> >>> - write_status_file();
>> >>> -
>> >>> - pool_shmem_exit(code);
>> >>> - exit(code);
>> >>> -}
>> >>> -
>> >>> -void notice_backend_error(int node_id)
>> >>> -{
>> >>> - int n = node_id;
>> >>> -
>> >>> - if (getpid() == mypid)
>> >>> - {
>> >>> - pool_log("notice_backend_error: called from pgpool
>> main.
>> >>> ignored.");
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - degenerate_backend_set(&n, 1);
>> >>> - }
>> >>> -}
>> >>> -
>> >>> -/* notice backend connection error using SIGUSR1 */
>> >>> -void degenerate_backend_set(int *node_id_set, int count)
>> >>> -{
>> >>> - pid_t parent = getppid();
>> >>> - int i;
>> >>> - bool need_signal = false;
>> >>> -#ifdef HAVE_SIGPROCMASK
>> >>> - sigset_t oldmask;
>> >>> -#else
>> >>> - int oldmask;
>> >>> -#endif
>> >>> -
>> >>> - if (pool_config->parallel_mode)
>> >>> - {
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - POOL_SETMASK2(&BlockSig, &oldmask);
>> >>> - pool_semaphore_lock(REQUEST_INFO_SEM);
>> >>> - Req_info->kind = NODE_DOWN_REQUEST;
>> >>> - for (i = 0; i < count; i++)
>> >>> - {
>> >>> - if (node_id_set[i] < 0 || node_id_set[i] >=
>> >>> MAX_NUM_BACKENDS ||
>> >>> - !VALID_BACKEND(node_id_set[i]))
>> >>> - {
>> >>> - pool_log("degenerate_backend_set: node %d is
>> not
>> >>> valid backend.", i);
>> >>> - continue;
>> >>> - }
>> >>> -
>> >>> - if
>> >>> (POOL_DISALLOW_TO_FAILOVER(BACKEND_INFO(node_id_set[i]).flag))
>> >>> - {
>> >>> - pool_log("degenerate_backend_set: %d failover
>> >>> request from pid %d is canceled because failover is disallowed",
>> >>> node_id_set[i], getpid());
>> >>> - continue;
>> >>> - }
>> >>> -
>> >>> - pool_log("degenerate_backend_set: %d fail over request
>> >>> from pid %d", node_id_set[i], getpid());
>> >>> - Req_info->node_id[i] = node_id_set[i];
>> >>> - need_signal = true;
>> >>> - }
>> >>> -
>> >>> - if (need_signal)
>> >>> - {
>> >>> - if (!pool_config->use_watchdog || WD_OK ==
>> >>> wd_degenerate_backend_set(node_id_set, count))
>> >>> - {
>> >>> - kill(parent, SIGUSR1);
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - pool_log("degenerate_backend_set: failover
>> >>> request from pid %d is canceled by other pgpool", getpid());
>> >>> - memset(Req_info->node_id, -1, sizeof(int) *
>> >>> MAX_NUM_BACKENDS);
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - pool_semaphore_unlock(REQUEST_INFO_SEM);
>> >>> - POOL_SETMASK(&oldmask);
>> >>> -}
>> >>> -
>> >>> -/* send promote node request using SIGUSR1 */
>> >>> -void promote_backend(int node_id)
>> >>> -{
>> >>> - pid_t parent = getppid();
>> >>> -
>> >>> - if (!MASTER_SLAVE || strcmp(pool_config->master_slave_sub_mode,
>> >>> MODE_STREAMREP))
>> >>> - {
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - if (node_id < 0 || node_id >= MAX_NUM_BACKENDS ||
>> >>> !VALID_BACKEND(node_id))
>> >>> - {
>> >>> - pool_error("promote_backend: node %d is not valid
>> >>> backend.", node_id);
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - pool_semaphore_lock(REQUEST_INFO_SEM);
>> >>> - Req_info->kind = PROMOTE_NODE_REQUEST;
>> >>> - Req_info->node_id[0] = node_id;
>> >>> - pool_log("promote_backend: %d promote node request from pid
>> %d",
>> >>> node_id, getpid());
>> >>> -
>> >>> - if (!pool_config->use_watchdog || WD_OK ==
>> >>> wd_promote_backend(node_id))
>> >>> - {
>> >>> - kill(parent, SIGUSR1);
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - pool_log("promote_backend: promote request from pid %d
>> is
>> >>> canceled by other pgpool", getpid());
>> >>> - Req_info->node_id[0] = -1;
>> >>> - }
>> >>> -
>> >>> - pool_semaphore_unlock(REQUEST_INFO_SEM);
>> >>> -}
>> >>> -
>> >>> -/* send failback request using SIGUSR1 */
>> >>> -void send_failback_request(int node_id)
>> >>> -{
>> >>> - pid_t parent = getppid();
>> >>> -
>> >>> - pool_log("send_failback_request: fail back %d th node request
>> >>> from pid %d", node_id, getpid());
>> >>> - Req_info->kind = NODE_UP_REQUEST;
>> >>> - Req_info->node_id[0] = node_id;
>> >>> -
>> >>> - if (node_id < 0 || node_id >= MAX_NUM_BACKENDS ||
>> >>> - (RAW_MODE && BACKEND_INFO(node_id).backend_status !=
>> >>> CON_DOWN && VALID_BACKEND(node_id)))
>> >>> - {
>> >>> - pool_error("send_failback_request: node %d is alive.",
>> >>> node_id);
>> >>> - Req_info->node_id[0] = -1;
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - if (pool_config->use_watchdog && WD_OK !=
>> >>> wd_send_failback_request(node_id))
>> >>> - {
>> >>> - pool_log("send_failback_request: failback request from
>> >>> pid %d is canceled by other pgpool", getpid());
>> >>> - Req_info->node_id[0] = -1;
>> >>> - return;
>> >>> - }
>> >>> - kill(parent, SIGUSR1);
>> >>> -}
>> >>> -
>> >>> -static RETSIGTYPE exit_handler(int sig)
>> >>> -{
>> >>> - int i;
>> >>> -
>> >>> - POOL_SETMASK(&AuthBlockSig);
>> >>> -
>> >>> - /*
>> >>> - * this could happen in a child process if a signal has been
>> sent
>> >>> - * before resetting signal handler
>> >>> - */
>> >>> - if (getpid() != mypid)
>> >>> - {
>> >>> - pool_debug("exit_handler: I am not parent");
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> - pool_shmem_exit(0);
>> >>> - exit(0);
>> >>> - }
>> >>> -
>> >>> - if (sig == SIGTERM)
>> >>> - pool_log("received smart shutdown request");
>> >>> - else if (sig == SIGINT)
>> >>> - pool_log("received fast shutdown request");
>> >>> - else if (sig == SIGQUIT)
>> >>> - pool_log("received immediate shutdown request");
>> >>> - else
>> >>> - {
>> >>> - pool_error("exit_handler: unknown signal received %d",
>> >>> sig);
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - exiting = 1;
>> >>> -
>> >>> - for (i = 0; i < pool_config->num_init_children; i++)
>> >>> - {
>> >>> - pid_t pid = process_info[i].pid;
>> >>> - if (pid)
>> >>> - {
>> >>> - kill(pid, sig);
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - kill(pcp_pid, sig);
>> >>> - kill(worker_pid, sig);
>> >>> -
>> >>> - if (pool_config->use_watchdog)
>> >>> - {
>> >>> - wd_kill_watchdog(sig);
>> >>> - }
>> >>> -
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> -
>> >>> - while (wait(NULL) > 0)
>> >>> - ;
>> >>> -
>> >>> - if (errno != ECHILD)
>> >>> - pool_error("wait() failed. reason:%s",
>> strerror(errno));
>> >>> -
>> >>> - process_info = NULL;
>> >>> - myexit(0);
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * Calculate next valid master node id.
>> >>> - * If no valid node found, returns -1.
>> >>> - */
>> >>> -static int get_next_master_node(void)
>> >>> -{
>> >>> - int i;
>> >>> -
>> >>> - for (i=0;i<pool_config->backend_desc->num_backends;i++)
>> >>> - {
>> >>> - /*
>> >>> - * Do not use VALID_BACKEND macro in raw mode.
>> >>> - * VALID_BACKEND return true only if the argument is
>> >>> master
>> >>> - * node id. In other words, standby nodes are false. So
>> >>> need
>> >>> - * to check backend status with VALID_BACKEND_RAW.
>> >>> - */
>> >>> - if (RAW_MODE)
>> >>> - {
>> >>> - if (VALID_BACKEND_RAW(i))
>> >>> - break;
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - if (VALID_BACKEND(i))
>> >>> - break;
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - if (i == pool_config->backend_desc->num_backends)
>> >>> - i = -1;
>> >>> -
>> >>> - return i;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * handle SIGUSR1
>> >>> - *
>> >>> - */
>> >>> -static RETSIGTYPE failover_handler(int sig)
>> >>> -{
>> >>> - POOL_SETMASK(&BlockSig);
>> >>> - failover_request = 1;
>> >>> - write(pipe_fds[1], "\0", 1);
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * backend connection error, failover/failback request, if possible
>> >>> - * failover() must be called under protecting signals.
>> >>> - */
>> >>> -static void failover(void)
>> >>> -{
>> >>> - int i;
>> >>> - int node_id;
>> >>> - bool by_health_check;
>> >>> - int new_master;
>> >>> - int new_primary;
>> >>> - int nodes[MAX_NUM_BACKENDS];
>> >>> - bool need_to_restart_children;
>> >>> - int status;
>> >>> - int sts;
>> >>> -
>> >>> - pool_debug("failover_handler called");
>> >>> -
>> >>> - memset(nodes, 0, sizeof(int) * MAX_NUM_BACKENDS);
>> >>> -
>> >>> - /*
>> >>> - * this could happen in a child process if a signal has been
>> sent
>> >>> - * before resetting signal handler
>> >>> - */
>> >>> - if (getpid() != mypid)
>> >>> - {
>> >>> - pool_debug("failover_handler: I am not parent");
>> >>> - kill(pcp_pid, SIGUSR2);
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - /*
>> >>> - * processing SIGTERM, SIGINT or SIGQUIT
>> >>> - */
>> >>> - if (exiting)
>> >>> - {
>> >>> - pool_debug("failover_handler called while exiting");
>> >>> - kill(pcp_pid, SIGUSR2);
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - /*
>> >>> - * processing fail over or switch over
>> >>> - */
>> >>> - if (switching)
>> >>> - {
>> >>> - pool_debug("failover_handler called while switching");
>> >>> - kill(pcp_pid, SIGUSR2);
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - pool_semaphore_lock(REQUEST_INFO_SEM);
>> >>> -
>> >>> - if (Req_info->kind == CLOSE_IDLE_REQUEST)
>> >>> - {
>> >>> - pool_semaphore_unlock(REQUEST_INFO_SEM);
>> >>> - kill_all_children(SIGUSR1);
>> >>> - kill(pcp_pid, SIGUSR2);
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - /*
>> >>> - * if not in replication mode/master slave mode, we treat this
>> a
>> >>> restart request.
>> >>> - * otherwise we need to check if we have already failovered.
>> >>> - */
>> >>> - pool_debug("failover_handler: starting to select new master
>> >>> node");
>> >>> - switching = 1;
>> >>> - Req_info->switching = true;
>> >>> - node_id = Req_info->node_id[0];
>> >>> -
>> >>> - /* start of command inter-lock with watchdog */
>> >>> - if (pool_config->use_watchdog)
>> >>> - {
>> >>> - by_health_check = (!failover_request &&
>> >>> Req_info->kind==NODE_DOWN_REQUEST);
>> >>> - wd_start_interlock(by_health_check);
>> >>> - }
>> >>> -
>> >>> - /* failback request? */
>> >>> - if (Req_info->kind == NODE_UP_REQUEST)
>> >>> - {
>> >>> - if (node_id >= MAX_NUM_BACKENDS ||
>> >>> - (Req_info->kind == NODE_UP_REQUEST &&
>> !(RAW_MODE
>> >>> &&
>> >>> - BACKEND_INFO(node_id).backend_status == CON_DOWN) &&
>> >>> VALID_BACKEND(node_id)) ||
>> >>> - (Req_info->kind == NODE_DOWN_REQUEST &&
>> >>> !VALID_BACKEND(node_id)))
>> >>> - {
>> >>> - pool_semaphore_unlock(REQUEST_INFO_SEM);
>> >>> - pool_error("failover_handler: invalid node_id
>> %d
>> >>> status:%d MAX_NUM_BACKENDS: %d", node_id,
>> >>> -
>> >>> BACKEND_INFO(node_id).backend_status, MAX_NUM_BACKENDS);
>> >>> - kill(pcp_pid, SIGUSR2);
>> >>> - switching = 0;
>> >>> - Req_info->switching = false;
>> >>> -
>> >>> - /* end of command inter-lock */
>> >>> - if (pool_config->use_watchdog)
>> >>> - wd_leave_interlock();
>> >>> -
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - pool_log("starting fail back. reconnect host %s(%d)",
>> >>> -
>> BACKEND_INFO(node_id).backend_hostname,
>> >>> - BACKEND_INFO(node_id).backend_port);
>> >>> - BACKEND_INFO(node_id).backend_status =
>> CON_CONNECT_WAIT;
>> >>> /* unset down status */
>> >>> -
>> >>> - /* wait for failback command lock or to be lock holder
>> */
>> >>> - if (pool_config->use_watchdog &&
>> !wd_am_I_lock_holder())
>> >>> - {
>> >>> - wd_wait_for_lock(WD_FAILBACK_COMMAND_LOCK);
>> >>> - }
>> >>> - /* execute failback command if lock holder */
>> >>> - if (!pool_config->use_watchdog ||
>> wd_am_I_lock_holder())
>> >>> - {
>> >>> - trigger_failover_command(node_id,
>> >>> pool_config->failback_command,
>> >>> -
>> >>> MASTER_NODE_ID, get_next_master_node(), PRIMARY_NODE_ID);
>> >>> -
>> >>> - /* unlock failback command */
>> >>> - if (pool_config->use_watchdog)
>> >>> - wd_unlock(WD_FAILBACK_COMMAND_LOCK);
>> >>> - }
>> >>> - }
>> >>> - else if (Req_info->kind == PROMOTE_NODE_REQUEST)
>> >>> - {
>> >>> - if (node_id != -1 && VALID_BACKEND(node_id))
>> >>> - {
>> >>> - pool_log("starting promotion. promote host
>> >>> %s(%d)",
>> >>> -
>> >>> BACKEND_INFO(node_id).backend_hostname,
>> >>> -
>> >>> BACKEND_INFO(node_id).backend_port);
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - pool_log("failover: no backends are promoted");
>> >>> - pool_semaphore_unlock(REQUEST_INFO_SEM);
>> >>> - kill(pcp_pid, SIGUSR2);
>> >>> - switching = 0;
>> >>> - Req_info->switching = false;
>> >>> -
>> >>> - /* end of command inter-lock */
>> >>> - if (pool_config->use_watchdog)
>> >>> - wd_leave_interlock();
>> >>> -
>> >>> - return;
>> >>> - }
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - int cnt = 0;
>> >>> -
>> >>> - for (i = 0; i < MAX_NUM_BACKENDS; i++)
>> >>> - {
>> >>> - if (Req_info->node_id[i] != -1 &&
>> >>> - ((RAW_MODE &&
>> >>> VALID_BACKEND_RAW(Req_info->node_id[i])) ||
>> >>> - VALID_BACKEND(Req_info->node_id[i])))
>> >>> - {
>> >>> - pool_log("starting degeneration.
>> shutdown
>> >>> host %s(%d)",
>> >>> -
>> >>> BACKEND_INFO(Req_info->node_id[i]).backend_hostname,
>> >>> -
>> >>> BACKEND_INFO(Req_info->node_id[i]).backend_port);
>> >>> -
>> >>> -
>> >>> BACKEND_INFO(Req_info->node_id[i]).backend_status = CON_DOWN; /* set
>> down
>> >>> status */
>> >>> - /* save down node */
>> >>> - nodes[Req_info->node_id[i]] = 1;
>> >>> - cnt++;
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - if (cnt == 0)
>> >>> - {
>> >>> - pool_log("failover: no backends are
>> degenerated");
>> >>> - pool_semaphore_unlock(REQUEST_INFO_SEM);
>> >>> - kill(pcp_pid, SIGUSR2);
>> >>> - switching = 0;
>> >>> - Req_info->switching = false;
>> >>> -
>> >>> - /* end of command inter-lock */
>> >>> - if (pool_config->use_watchdog)
>> >>> - wd_leave_interlock();
>> >>> -
>> >>> - return;
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - new_master = get_next_master_node();
>> >>> -
>> >>> - if (new_master < 0)
>> >>> - {
>> >>> - pool_error("failover_handler: no valid DB node found");
>> >>> - }
>> >>> -
>> >>> -/*
>> >>> - * Before we tried to minimize restarting pgpool to protect existing
>> >>> - * connections from clients to pgpool children. What we did here was,
>> >>> - * if children other than master went down, we did not fail over.
>> >>> - * This is wrong. Think about following scenario. If someone
>> >>> - * accidentally plugs out the network cable, the TCP/IP stack keeps
>> >>> - * retrying for long time (typically 2 hours). The only way to stop
>> >>> - * the retry is restarting the process. Bottom line is, we need to
>> >>> - * restart all children in any case. See pgpool-general list posting
>> >>> - * "TCP connections are *not* closed when a backend timeout" on Jul 13
>> >>> - * 2008 for more details.
>> >>> - */
>> >>> -#ifdef NOT_USED
>> >>> - else
>> >>> - {
>> >>> - if (Req_info->master_node_id == new_master &&
>> *InRecovery
>> >>> == RECOVERY_INIT)
>> >>> - {
>> >>> - pool_log("failover_handler: do not restart
>> >>> pgpool. same master node %d was selected", new_master);
>> >>> - if (Req_info->kind == NODE_UP_REQUEST)
>> >>> - {
>> >>> - pool_log("failback done. reconnect host
>> >>> %s(%d)",
>> >>> -
>> >>> BACKEND_INFO(node_id).backend_hostname,
>> >>> -
>> >>> BACKEND_INFO(node_id).backend_port);
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - pool_log("failover done. shutdown host
>> >>> %s(%d)",
>> >>> -
>> >>> BACKEND_INFO(node_id).backend_hostname,
>> >>> -
>> >>> BACKEND_INFO(node_id).backend_port);
>> >>> - }
>> >>> -
>> >>> - /* exec failover_command */
>> >>> - for (i = 0; i <
>> >>> pool_config->backend_desc->num_backends; i++)
>> >>> - {
>> >>> - if (nodes[i])
>> >>> - trigger_failover_command(i,
>> >>> pool_config->failover_command);
>> >>> - }
>> >>> -
>> >>> - pool_semaphore_unlock(REQUEST_INFO_SEM);
>> >>> - switching = 0;
>> >>> - Req_info->switching = false;
>> >>> - kill(pcp_pid, SIGUSR2);
>> >>> - switching = 0;
>> >>> - Req_info->switching = false;
>> >>> - return;
>> >>> - }
>> >>> - }
>> >>> -#endif
>> >>> -
>> >>> -
>> >>> - /* On 2011/5/2 Tatsuo Ishii says: if mode is streaming replication
>> >>> - * and request is NODE_UP_REQUEST(failback case) we don't need
>> to
>> >>> - * restart all children. Existing session will not use newly
>> >>> - * attached node, but load balanced node is not changed until
>> this
>> >>> - * session ends, so it's harmless anyway.
>> >>> - */
>> >>> - if (MASTER_SLAVE && !strcmp(pool_config->master_slave_sub_mode,
>> >>> MODE_STREAMREP) &&
>> >>> - Req_info->kind == NODE_UP_REQUEST)
>> >>> - {
>> >>> - pool_log("Do not restart children because we are
>> >>> failbacking node id %d host%s port:%d and we are in streaming
>> replication
>> >>> mode", node_id,
>> >>> -
>> BACKEND_INFO(node_id).backend_hostname,
>> >>> - BACKEND_INFO(node_id).backend_port);
>> >>> -
>> >>> - need_to_restart_children = false;
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - pool_log("Restart all children");
>> >>> -
>> >>> - /* kill all children */
>> >>> - for (i = 0; i < pool_config->num_init_children; i++)
>> >>> - {
>> >>> - pid_t pid = process_info[i].pid;
>> >>> - if (pid)
>> >>> - {
>> >>> - kill(pid, SIGQUIT);
>> >>> - pool_debug("failover_handler: kill %d",
>> >>> pid);
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - need_to_restart_children = true;
>> >>> - }
>> >>> -
>> >>> - /* wait for failover command lock or to be lock holder*/
>> >>> - if (pool_config->use_watchdog && !wd_am_I_lock_holder())
>> >>> - {
>> >>> - wd_wait_for_lock(WD_FAILOVER_COMMAND_LOCK);
>> >>> - }
>> >>> -
>> >>> - /* execute failover command if lock holder */
>> >>> - if (!pool_config->use_watchdog || wd_am_I_lock_holder())
>> >>> - {
>> >>> - /* Exec failover_command if needed */
>> >>> - for (i = 0; i <
>> pool_config->backend_desc->num_backends;
>> >>> i++)
>> >>> - {
>> >>> - if (nodes[i])
>> >>> - trigger_failover_command(i,
>> >>> pool_config->failover_command,
>> >>> -
>> >>> MASTER_NODE_ID, new_master, PRIMARY_NODE_ID);
>> >>> - }
>> >>> -
>> >>> - /* unlock failover command */
>> >>> - if (pool_config->use_watchdog)
>> >>> - wd_unlock(WD_FAILOVER_COMMAND_LOCK);
>> >>> - }
>> >>> -
>> >>> -
>> >>> -/* no need to wait since it will be done in reap_handler */
>> >>> -#ifdef NOT_USED
>> >>> - while (wait(NULL) > 0)
>> >>> - ;
>> >>> -
>> >>> - if (errno != ECHILD)
>> >>> - pool_error("failover_handler: wait() failed.
>> reason:%s",
>> >>> strerror(errno));
>> >>> -#endif
>> >>> -
>> >>> - if (Req_info->kind == PROMOTE_NODE_REQUEST &&
>> >>> VALID_BACKEND(node_id))
>> >>> - new_primary = node_id;
>> >>> - else
>> >>> - new_primary = find_primary_node_repeatedly();
>> >>> -
>> >>> - /*
>> >>> - * If follow_master_command is provided and in master/slave
>> >>> - * streaming replication mode, we start degenerating all
>> backends
>> >>> - * as they are not replicated anymore.
>> >>> - */
>> >>> - int follow_cnt = 0;
>> >>> - if (MASTER_SLAVE && !strcmp(pool_config->master_slave_sub_mode,
>> >>> MODE_STREAMREP))
>> >>> - {
>> >>> - if (*pool_config->follow_master_command != '\0' ||
>> >>> - Req_info->kind == PROMOTE_NODE_REQUEST)
>> >>> - {
>> >>> - /* only if the failover is against the current
>> >>> primary */
>> >>> - if (((Req_info->kind == NODE_DOWN_REQUEST) &&
>> >>> - (nodes[Req_info->primary_node_id])) ||
>> >>> - ((Req_info->kind ==
>> PROMOTE_NODE_REQUEST)
>> >>> &&
>> >>> - (VALID_BACKEND(node_id)))) {
>> >>> -
>> >>> - for (i = 0; i <
>> >>> pool_config->backend_desc->num_backends; i++)
>> >>> - {
>> >>> - /* do not degenerate the new
>> >>> primary */
>> >>> - if ((new_primary >= 0) && (i !=
>> >>> new_primary)) {
>> >>> - BackendInfo *bkinfo;
>> >>> - bkinfo =
>> >>> pool_get_node_info(i);
>> >>> - pool_log("starting
>> follow
>> >>> degeneration. shutdown host %s(%d)",
>> >>> -
>> >>> bkinfo->backend_hostname,
>> >>> -
>> >>> bkinfo->backend_port);
>> >>> - bkinfo->backend_status
>> =
>> >>> CON_DOWN; /* set down status */
>> >>> - follow_cnt++;
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - if (follow_cnt == 0)
>> >>> - {
>> >>> - pool_log("failover: no follow
>> >>> backends are degenerated");
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - /* update new master node */
>> >>> - new_master =
>> >>> get_next_master_node();
>> >>> - pool_log("failover: %d follow
>> >>> backends have been degenerated", follow_cnt);
>> >>> - }
>> >>> - }
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - memset(Req_info->node_id, -1, sizeof(int) * MAX_NUM_BACKENDS);
>> >>> - pool_semaphore_unlock(REQUEST_INFO_SEM);
>> >>> -
>> >>> - /* wait for follow_master_command lock or to be lock holder */
>> >>> - if (pool_config->use_watchdog && !wd_am_I_lock_holder())
>> >>> - {
>> >>> - wd_wait_for_lock(WD_FOLLOW_MASTER_COMMAND_LOCK);
>> >>> - }
>> >>> -
>> >>> - /* execute follow_master_command */
>> >>> - if (!pool_config->use_watchdog || wd_am_I_lock_holder())
>> >>> - {
>> >>> - if ((follow_cnt > 0) &&
>> >>> (*pool_config->follow_master_command != '\0'))
>> >>> - {
>> >>> - follow_pid =
>> >>> fork_follow_child(Req_info->master_node_id, new_primary,
>> >>> -
>> >>> Req_info->primary_node_id);
>> >>> - }
>> >>> -
>> >>> - /* unlock follow_master_command */
>> >>> - if (pool_config->use_watchdog)
>> >>> - wd_unlock(WD_FOLLOW_MASTER_COMMAND_LOCK);
>> >>> - }
>> >>> -
>> >>> - /* end of command inter-lock */
>> >>> - if (pool_config->use_watchdog)
>> >>> - wd_end_interlock();
>> >>> -
>> >>> - /* Save primary node id */
>> >>> - Req_info->primary_node_id = new_primary;
>> >>> - pool_log("failover: set new primary node: %d",
>> >>> Req_info->primary_node_id);
>> >>> -
>> >>> - if (new_master >= 0)
>> >>> - {
>> >>> - Req_info->master_node_id = new_master;
>> >>> - pool_log("failover: set new master node: %d",
>> >>> Req_info->master_node_id);
>> >>> - }
>> >>> -
>> >>> -
>> >>> - /* Fork the children if needed */
>> >>> - if (need_to_restart_children)
>> >>> - {
>> >>> - for (i=0;i<pool_config->num_init_children;i++)
>> >>> - {
>> >>> -
>> >>> - /*
>> >>> - * Try to kill pgpool child because previous
>> kill
>> >>> signal
>> >>> - * may not be received by pgpool child. This
>> >>> could happen
>> >>> - * if multiple PostgreSQL are going down (or
>> even
>> >>> starting
>> >>> - * pgpool, without starting PostgreSQL can
>> >>> trigger this).
>> >>> - * Child calls degenerate_backend() and it
>> tries
>> >>> to aquire
>> >>> - * semaphore to write a failover request. In
>> this
>> >>> case the
>> >>> - * signal mask is set as well, thus signals are
>> >>> never
>> >>> - * received.
>> >>> - */
>> >>> - kill(process_info[i].pid, SIGQUIT);
>> >>> -
>> >>> - process_info[i].pid = fork_a_child(unix_fd,
>> >>> inet_fd, i);
>> >>> - process_info[i].start_time = time(NULL);
>> >>> - }
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - /* Set restart request to each child. Children will
>> >>> exit(1)
>> >>> - * whenever they are idle to restart.
>> >>> - */
>> >>> - for (i=0;i<pool_config->num_init_children;i++)
>> >>> - {
>> >>> - process_info[i].need_to_restart = 1;
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - /*
>> >>> - * Send restart request to worker child.
>> >>> - */
>> >>> - kill(worker_pid, SIGUSR1);
>> >>> -
>> >>> - if (Req_info->kind == NODE_UP_REQUEST)
>> >>> - {
>> >>> - pool_log("failback done. reconnect host %s(%d)",
>> >>> -
>> BACKEND_INFO(node_id).backend_hostname,
>> >>> - BACKEND_INFO(node_id).backend_port);
>> >>> - }
>> >>> - else if (Req_info->kind == PROMOTE_NODE_REQUEST)
>> >>> - {
>> >>> - pool_log("promotion done. promoted host %s(%d)",
>> >>> -
>> BACKEND_INFO(node_id).backend_hostname,
>> >>> - BACKEND_INFO(node_id).backend_port);
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - pool_log("failover done. shutdown host %s(%d)",
>> >>> -
>> BACKEND_INFO(node_id).backend_hostname,
>> >>> - BACKEND_INFO(node_id).backend_port);
>> >>> - }
>> >>> -
>> >>> - switching = 0;
>> >>> - Req_info->switching = false;
>> >>> -
>> >>> - /* kick wakeup_handler in pcp_child to notice that
>> >>> - * failover/failback done
>> >>> - */
>> >>> - kill(pcp_pid, SIGUSR2);
>> >>> -
>> >>> - sleep(1);
>> >>> -
>> >>> - /*
>> >>> - * Send restart request to pcp child.
>> >>> - */
>> >>> - kill(pcp_pid, SIGUSR1);
>> >>> - for (;;)
>> >>> - {
>> >>> - sts = waitpid(pcp_pid, &status, 0);
>> >>> - if (sts != -1)
>> >>> - break;
>> >>> - if (sts == -1)
>> >>> - {
>> >>> - if (errno == EINTR)
>> >>> - continue;
>> >>> - else
>> >>> - {
>> >>> - pool_error("failover: waitpid failed.
>> >>> reason: %s", strerror(errno));
>> >>> - return;
>> >>> - }
>> >>> - }
>> >>> - }
>> >>> - if (WIFSIGNALED(status))
>> >>> - pool_log("PCP child %d exits with status %d by signal
>> %d
>> >>> in failover()", pcp_pid, status, WTERMSIG(status));
>> >>> - else
>> >>> - pool_log("PCP child %d exits with status %d in
>> >>> failover()", pcp_pid, status);
>> >>> -
>> >>> - pcp_pid = pcp_fork_a_child(pcp_unix_fd, pcp_inet_fd,
>> >>> pcp_conf_file);
>> >>> - pool_log("fork a new PCP child pid %d in failover()", pcp_pid);
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * health check timer handler
>> >>> - */
>> >>> -static RETSIGTYPE health_check_timer_handler(int sig)
>> >>> -{
>> >>> - POOL_SETMASK(&BlockSig);
>> >>> - health_check_timer_expired = 1;
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> -}
>> >>> -
>> >>> -
>> >>> -/*
>> >>> - * Check if we can connect to the backend
>> >>> - * returns 0 for OK. otherwise returns backend id + 1
>> >>> - */
>> >>> -static int health_check(void)
>> >>> -{
>> >>> - POOL_CONNECTION_POOL_SLOT *slot;
>> >>> - BackendInfo *bkinfo;
>> >>> - static bool is_first = true;
>> >>> - static char *dbname;
>> >>> - int i;
>> >>> -
>> >>> - /* Do not execute health check during recovery */
>> >>> - if (*InRecovery)
>> >>> - return 0;
>> >>> -
>> >>> - Retry:
>> >>> - /*
>> >>> - * First we try with "postgres" database.
>> >>> - */
>> >>> - if (is_first)
>> >>> - dbname = "postgres";
>> >>> -
>> >>> - for (i=0;i<pool_config->backend_desc->num_backends;i++)
>> >>> - {
>> >>> - /*
>> >>> - * Make sure that health check timer has not been
>> expired.
>> >>> - * Before called health_check(),
>> >>> health_check_timer_expired is
>> >>> - * set to 0. However it is possible that while
>> >>> processing DB
>> >>> - * nodes health check timer expired.
>> >>> - */
>> >>> - if (health_check_timer_expired)
>> >>> - {
>> >>> - pool_log("health_check: health check timer has
>> >>> been already expired before attempting to connect to %d th backend",
>> i);
>> >>> - return i+1;
>> >>> - }
>> >>> -
>> >>> - bkinfo = pool_get_node_info(i);
>> >>> -
>> >>> - pool_debug("health_check: %d th DB node status: %d", i,
>> >>> bkinfo->backend_status);
>> >>> -
>> >>> - if (bkinfo->backend_status == CON_UNUSED ||
>> >>> - bkinfo->backend_status == CON_DOWN)
>> >>> - continue;
>> >>> -
>> >>> - slot =
>> >>> make_persistent_db_connection(bkinfo->backend_hostname,
>> >>> -
>> >>> bkinfo->backend_port,
>> >>> -
>> >>> dbname,
>> >>> -
>> >>> pool_config->health_check_user,
>> >>> -
>> >>> pool_config->health_check_password, false);
>> >>> -
>> >>> - if (is_first)
>> >>> - is_first = false;
>> >>> -
>> >>> - if (!slot)
>> >>> - {
>> >>> - /*
>> >>> - * Retry with template1 unless health check
>> timer
>> >>> is expired.
>> >>> - */
>> >>> - if (!strcmp(dbname, "postgres") &&
>> >>> health_check_timer_expired == 0)
>> >>> - {
>> >>> - dbname = "template1";
>> >>> - goto Retry;
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - pool_error("health check failed. %d th
>> >>> host %s at port %d is down",
>> >>> - i,
>> >>> -
>> >>> bkinfo->backend_hostname,
>> >>> -
>> bkinfo->backend_port);
>> >>> - return i+1;
>> >>> - }
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - discard_persistent_db_connection(slot);
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - return 0;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * check if we can connect to the SystemDB
>> >>> - * returns 0 for OK. otherwise returns -1
>> >>> - */
>> >>> -static int
>> >>> -system_db_health_check(void)
>> >>> -{
>> >>> - int fd;
>> >>> -
>> >>> - /* V2 startup packet */
>> >>> - typedef struct {
>> >>> - int len; /* startup packet length */
>> >>> - StartupPacket_v2 sp;
>> >>> - } MySp;
>> >>> - MySp mysp;
>> >>> - char kind;
>> >>> -
>> >>> - memset(&mysp, 0, sizeof(mysp));
>> >>> - mysp.len = htonl(296);
>> >>> - mysp.sp.protoVersion = htonl(PROTO_MAJOR_V2 << 16);
>> >>> - strcpy(mysp.sp.database, "template1");
>> >>> - strncpy(mysp.sp.user, SYSDB_INFO->user, sizeof(mysp.sp.user) -
>> 1);
>> >>> - *mysp.sp.options = '\0';
>> >>> - *mysp.sp.unused = '\0';
>> >>> - *mysp.sp.tty = '\0';
>> >>> -
>> >>> - pool_debug("health_check: SystemDB status: %d", SYSDB_STATUS);
>> >>> -
>> >>> - /* if SystemDB is already down, ignore */
>> >>> - if (SYSDB_STATUS == CON_UNUSED || SYSDB_STATUS == CON_DOWN)
>> >>> - return 0;
>> >>> -
>> >>> - if (*SYSDB_INFO->hostname == '/')
>> >>> - fd =
>> connect_unix_domain_socket_by_port(SYSDB_INFO->port,
>> >>> SYSDB_INFO->hostname, FALSE);
>> >>> - else
>> >>> - fd =
>> >>> connect_inet_domain_socket_by_port(SYSDB_INFO->hostname,
>> SYSDB_INFO->port,
>> >>> FALSE);
>> >>> -
>> >>> - if (fd < 0)
>> >>> - {
>> >>> - pool_error("health check failed. SystemDB host %s at
>> port
>> >>> %d is down",
>> >>> - SYSDB_INFO->hostname,
>> >>> - SYSDB_INFO->port);
>> >>> -
>> >>> - return -1;
>> >>> - }
>> >>> -
>> >>> - if (write(fd, &mysp, sizeof(mysp)) < 0)
>> >>> - {
>> >>> - pool_error("health check failed during write. SystemDB
>> >>> host %s at port %d is down",
>> >>> - SYSDB_INFO->hostname,
>> >>> - SYSDB_INFO->port);
>> >>> - close(fd);
>> >>> - return -1;
>> >>> - }
>> >>> -
>> >>> - read(fd, &kind, 1);
>> >>> -
>> >>> - if (write(fd, "X", 1) < 0)
>> >>> - {
>> >>> - pool_error("health check failed during write. SystemDB
>> >>> host %s at port %d is down",
>> >>> - SYSDB_INFO->hostname,
>> >>> - SYSDB_INFO->port);
>> >>> - close(fd);
>> >>> - return -1;
>> >>> - }
>> >>> -
>> >>> - close(fd);
>> >>> - return 0;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * handle SIGCHLD
>> >>> - */
>> >>> -static RETSIGTYPE reap_handler(int sig)
>> >>> -{
>> >>> - POOL_SETMASK(&BlockSig);
>> >>> - sigchld_request = 1;
>> >>> - write(pipe_fds[1], "\0", 1);
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * Attach zombie processes and restart child processes.
>> >>> - * reaper() must be called protected from signals.
>> >>> - */
>> >>> -static void reaper(void)
>> >>> -{
>> >>> - pid_t pid;
>> >>> - int status;
>> >>> - int i;
>> >>> -
>> >>> - pool_debug("reap_handler called");
>> >>> -
>> >>> - if (exiting)
>> >>> - {
>> >>> - pool_debug("reap_handler: exited due to exiting");
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - if (switching)
>> >>> - {
>> >>> - pool_debug("reap_handler: exited due to switching");
>> >>> - return;
>> >>> - }
>> >>> -
>> >>> - /* clear SIGCHLD request */
>> >>> - sigchld_request = 0;
>> >>> -
>> >>> -#ifdef HAVE_WAITPID
>> >>> - pool_debug("reap_handler: call waitpid");
>> >>> - while ((pid = waitpid(-1, &status, WNOHANG)) > 0)
>> >>> -#else
>> >>> - pool_debug("reap_handler: call wait3");
>> >>> - while ((pid = wait3(&status, WNOHANG, NULL)) > 0)
>> >>> -#endif
>> >>> - {
>> >>> - if (WIFSIGNALED(status) && WTERMSIG(status) == SIGSEGV)
>> >>> - {
>> >>> - /* Child terminated by segmentation fault.
>> Report
>> >>> it */
>> >>> - pool_error("Child process %d was terminated by
>> >>> segmentation fault", pid);
>> >>> - }
>> >>> -
>> >>> - /* if exiting child process was PCP handler */
>> >>> - if (pid == pcp_pid)
>> >>> - {
>> >>> - if (WIFSIGNALED(status))
>> >>> - pool_log("PCP child %d exits with
>> status
>> >>> %d by signal %d", pid, status, WTERMSIG(status));
>> >>> - else
>> >>> - pool_log("PCP child %d exits with
>> status
>> >>> %d", pid, status);
>> >>> -
>> >>> - pcp_pid = pcp_fork_a_child(pcp_unix_fd,
>> >>> pcp_inet_fd, pcp_conf_file);
>> >>> - pool_log("fork a new PCP child pid %d",
>> pcp_pid);
>> >>> - }
>> >>> -
>> >>> - /* exiting process was worker process */
>> >>> - else if (pid == worker_pid)
>> >>> - {
>> >>> - if (WIFSIGNALED(status))
>> >>> - pool_log("worker child %d exits with
>> >>> status %d by signal %d", pid, status, WTERMSIG(status));
>> >>> - else
>> >>> - pool_log("worker child %d exits with
>> >>> status %d", pid, status);
>> >>> -
>> >>> - if (status)
>> >>> - worker_pid = worker_fork_a_child();
>> >>> -
>> >>> - pool_log("fork a new worker child pid %d",
>> >>> worker_pid);
>> >>> - }
>> >>> -
>> >>> - /* exiting process was watchdog process */
>> >>> - else if (pool_config->use_watchdog &&
>> >>> wd_is_watchdog_pid(pid))
>> >>> - {
>> >>> - if (!wd_reaper_watchdog(pid, status))
>> >>> - {
>> >>> - pool_error("wd_reaper failed");
>> >>> - myexit(1);
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - else
>> >>> - {
>> >>> - if (WIFSIGNALED(status))
>> >>> - pool_debug("child %d exits with status
>> %d
>> >>> by signal %d", pid, status, WTERMSIG(status));
>> >>> - else
>> >>> - pool_debug("child %d exits with status
>> >>> %d", pid, status);
>> >>> -
>> >>> - /* look for exiting child's pid */
>> >>> - for (i=0;i<pool_config->num_init_children;i++)
>> >>> - {
>> >>> - if (pid == process_info[i].pid)
>> >>> - {
>> >>> - /* if found, fork a new child
>> */
>> >>> - if (!switching && !exiting &&
>> >>> status)
>> >>> - {
>> >>> - process_info[i].pid =
>> >>> fork_a_child(unix_fd, inet_fd, i);
>> >>> -
>> >>> process_info[i].start_time = time(NULL);
>> >>> - pool_debug("fork a new
>> >>> child pid %d", process_info[i].pid);
>> >>> - break;
>> >>> - }
>> >>> - }
>> >>> - }
>> >>> - }
>> >>> - }
>> >>> - pool_debug("reap_handler: normally exited");
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * get node information specified by node_number
>> >>> - */
>> >>> -BackendInfo *
>> >>> -pool_get_node_info(int node_number)
>> >>> -{
>> >>> - if (node_number < 0 || node_number >= NUM_BACKENDS)
>> >>> - return NULL;
>> >>> -
>> >>> - return &BACKEND_INFO(node_number);
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * get number of nodes
>> >>> - */
>> >>> -int
>> >>> -pool_get_node_count(void)
>> >>> -{
>> >>> - return NUM_BACKENDS;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * get process ids
>> >>> - */
>> >>> -int *
>> >>> -pool_get_process_list(int *array_size)
>> >>> -{
>> >>> - int *array;
>> >>> - int i;
>> >>> -
>> >>> - *array_size = pool_config->num_init_children;
>> >>> - array = calloc(*array_size, sizeof(int));
>> >>> - for (i = 0; i < *array_size; i++)
>> >>> - array[i] = process_info[i].pid;
>> >>> -
>> >>> - return array;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * get process information specified by pid
>> >>> - */
>> >>> -ProcessInfo *
>> >>> -pool_get_process_info(pid_t pid)
>> >>> -{
>> >>> - int i;
>> >>> -
>> >>> - for (i = 0; i < pool_config->num_init_children; i++)
>> >>> - if (process_info[i].pid == pid)
>> >>> - return &process_info[i];
>> >>> -
>> >>> - return NULL;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * get System DB information
>> >>> - */
>> >>> -SystemDBInfo *
>> >>> -pool_get_system_db_info(void)
>> >>> -{
>> >>> - if (system_db_info == NULL)
>> >>> - return NULL;
>> >>> -
>> >>> - return system_db_info->info;
>> >>> -}
>> >>> -
>> >>> -
>> >>> -/*
>> >>> - * handle SIGUSR2
>> >>> - * Wakeup all processes
>> >>> - */
>> >>> -static void wakeup_children(void)
>> >>> -{
>> >>> - kill_all_children(SIGUSR2);
>> >>> -}
>> >>> -
>> >>> -
>> >>> -static RETSIGTYPE wakeup_handler(int sig)
>> >>> -{
>> >>> - POOL_SETMASK(&BlockSig);
>> >>> - wakeup_request = 1;
>> >>> - write(pipe_fds[1], "\0", 1);
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * handle SIGHUP
>> >>> - *
>> >>> - */
>> >>> -static RETSIGTYPE reload_config_handler(int sig)
>> >>> -{
>> >>> - POOL_SETMASK(&BlockSig);
>> >>> - reload_config_request = 1;
>> >>> - write(pipe_fds[1], "\0", 1);
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> -}
>> >>> -
>> >>> -static void reload_config(void)
>> >>> -{
>> >>> - pool_log("reload config files.");
>> >>> - pool_get_config(conf_file, RELOAD_CONFIG);
>> >>> - if (pool_config->enable_pool_hba)
>> >>> - load_hba(hba_file);
>> >>> - if (pool_config->parallel_mode)
>> >>> - pool_memset_system_db_info(system_db_info->info);
>> >>> - kill_all_children(SIGHUP);
>> >>> -
>> >>> - if (worker_pid)
>> >>> - kill(worker_pid, SIGHUP);
>> >>> -}
>> >>> -
>> >>> -static void kill_all_children(int sig)
>> >>> -{
>> >>> - int i;
>> >>> -
>> >>> - /* kill all children */
>> >>> - for (i = 0; i < pool_config->num_init_children; i++)
>> >>> - {
>> >>> - pid_t pid = process_info[i].pid;
>> >>> - if (pid)
>> >>> - {
>> >>> - kill(pid, sig);
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - /* make PCP process reload as well */
>> >>> - if (sig == SIGHUP)
>> >>> - kill(pcp_pid, sig);
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * pause in a period specified by timeout. If any data is coming
>> >>> - * through pipe_fds[0], that means one of: failover request(SIGUSR1),
>> >>> - * SIGCHLD received, children wake up request(SIGUSR2 used in on line
>> >>> - * recovery processing) or config file reload request(SIGHUP) has been
>> >>> - * occurred. In this case this function returns 1.
>> >>> - * otherwise 0: (no signal event occurred), -1: (error)
>> >>> - * XXX: is it OK that select(2) error is ignored here?
>> >>> - */
>> >>> -static int pool_pause(struct timeval *timeout)
>> >>> -{
>> >>> - fd_set rfds;
>> >>> - int n;
>> >>> - char dummy;
>> >>> -
>> >>> - FD_ZERO(&rfds);
>> >>> - FD_SET(pipe_fds[0], &rfds);
>> >>> - n = select(pipe_fds[0]+1, &rfds, NULL, NULL, timeout);
>> >>> - if (n == 1)
>> >>> - read(pipe_fds[0], &dummy, 1);
>> >>> - return n;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * sleep for seconds specified by "second". Unlike pool_pause(), this
>> >>> - * function guarantees that it will sleep for specified seconds. This
>> >>> - * function uses pool_pause() internally. If it informs that there is
>> >>> - * a pending signal event, they are processed using CHECK_REQUEST
>> >>> - * macro. Note that most of these processes are done while all signals
>> >>> - * are blocked.
>> >>> - */
>> >>> -void pool_sleep(unsigned int second)
>> >>> -{
>> >>> - struct timeval current_time, sleep_time;
>> >>> -
>> >>> - gettimeofday(¤t_time, NULL);
>> >>> - sleep_time.tv_sec = second + current_time.tv_sec;
>> >>> - sleep_time.tv_usec = current_time.tv_usec;
>> >>> -
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> - while (sleep_time.tv_sec > current_time.tv_sec)
>> >>> - {
>> >>> - struct timeval timeout;
>> >>> - int r;
>> >>> -
>> >>> - timeout.tv_sec = sleep_time.tv_sec -
>> current_time.tv_sec;
>> >>> - timeout.tv_usec = sleep_time.tv_usec -
>> >>> current_time.tv_usec;
>> >>> - if (timeout.tv_usec < 0)
>> >>> - {
>> >>> - timeout.tv_sec--;
>> >>> - timeout.tv_usec += 1000000;
>> >>> - }
>> >>> -
>> >>> - r = pool_pause(&timeout);
>> >>> - POOL_SETMASK(&BlockSig);
>> >>> - if (r > 0)
>> >>> - CHECK_REQUEST;
>> >>> - POOL_SETMASK(&UnBlockSig);
>> >>> - gettimeofday(¤t_time, NULL);
>> >>> - }
>> >>> - POOL_SETMASK(&BlockSig);
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * get_config_file_name: return full path of pgpool.conf.
>> >>> - */
>> >>> -char *get_config_file_name(void)
>> >>> -{
>> >>> - return conf_file;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * get_config_file_name: return full path of pool_hba.conf.
>> >>> - */
>> >>> -char *get_hba_file_name(void)
>> >>> + * get_config_file_name: return full path of pool_hba.conf.
>> >>> + */
>> >>> +char *get_hba_file_name(void)
>> >>> {
>> >>> return hba_file;
>> >>> }
>> >>> -
>> >>> -/*
>> >>> - * trigger_failover_command: execute specified command at failover.
>> >>> - * command_line is null-terminated string.
>> >>> - */
>> >>> -static int trigger_failover_command(int node, const char
>> *command_line,
>> >>> -
>> >>> int old_master, int new_master, int old_primary)
>> >>> -{
>> >>> - int r = 0;
>> >>> - String *exec_cmd;
>> >>> - char port_buf[6];
>> >>> - char buf[2];
>> >>> - BackendInfo *info;
>> >>> - BackendInfo *newmaster;
>> >>> -
>> >>> - if (command_line == NULL || (strlen(command_line) == 0))
>> >>> - return 0;
>> >>> -
>> >>> - /* check failed nodeID */
>> >>> - if (node < 0 || node > NUM_BACKENDS)
>> >>> - return -1;
>> >>> -
>> >>> - info = pool_get_node_info(node);
>> >>> - if (!info)
>> >>> - return -1;
>> >>> -
>> >>> - buf[1] = '\0';
>> >>> - pool_memory = pool_memory_create(PREPARE_BLOCK_SIZE);
>> >>> - if (!pool_memory)
>> >>> - {
>> >>> - pool_error("trigger_failover_command:
>> >>> pool_memory_create() failed");
>> >>> - return -1;
>> >>> - }
>> >>> - exec_cmd = init_string("");
>> >>> -
>> >>> - while (*command_line)
>> >>> - {
>> >>> - if (*command_line == '%')
>> >>> - {
>> >>> - if (*(command_line + 1))
>> >>> - {
>> >>> - char val = *(command_line + 1);
>> >>> - switch (val)
>> >>> - {
>> >>> - case 'p': /* failed node port
>> */
>> >>> - snprintf(port_buf,
>> >>> sizeof(port_buf), "%d", info->backend_port);
>> >>> -
>> >>> string_append_char(exec_cmd, port_buf);
>> >>> - break;
>> >>> -
>> >>> - case 'D': /* failed node
>> database
>> >>> directory */
>> >>> -
>> >>> string_append_char(exec_cmd, info->backend_data_directory);
>> >>> - break;
>> >>> -
>> >>> - case 'd': /* failed node id */
>> >>> - snprintf(port_buf,
>> >>> sizeof(port_buf), "%d", node);
>> >>> -
>> >>> string_append_char(exec_cmd, port_buf);
>> >>> - break;
>> >>> -
>> >>> - case 'h': /* failed host name
>> */
>> >>> -
>> >>> string_append_char(exec_cmd, info->backend_hostname);
>> >>> - break;
>> >>> -
>> >>> - case 'H': /* new master host
>> name
>> >>> */
>> >>> - newmaster =
>> >>> pool_get_node_info(new_master);
>> >>> - if (newmaster)
>> >>> -
>> >>> string_append_char(exec_cmd, newmaster->backend_hostname);
>> >>> - else
>> >>> - /* no valid new
>> >>> master */
>> >>> -
>> >>> string_append_char(exec_cmd, "");
>> >>> - break;
>> >>> -
>> >>> - case 'm': /* new master node
>> id */
>> >>> - snprintf(port_buf,
>> >>> sizeof(port_buf), "%d", new_master);
>> >>> -
>> >>> string_append_char(exec_cmd, port_buf);
>> >>> - break;
>> >>> -
>> >>> - case 'r': /* new master port */
>> >>> - newmaster =
>> >>> pool_get_node_info(get_next_master_node());
>> >>> - if (newmaster)
>> >>> - {
>> >>> -
>> >>> snprintf(port_buf, sizeof(port_buf), "%d", newmaster->backend_port);
>> >>> -
>> >>> string_append_char(exec_cmd, port_buf);
>> >>> - }
>> >>> - else
>> >>> - /* no valid new
>> >>> master */
>> >>> -
>> >>> string_append_char(exec_cmd, "");
>> >>> - break;
>> >>> -
>> >>> - case 'R': /* new master
>> database
>> >>> directory */
>> >>> - newmaster =
>> >>> pool_get_node_info(get_next_master_node());
>> >>> - if (newmaster)
>> >>> -
>> >>> string_append_char(exec_cmd, newmaster->backend_data_directory);
>> >>> - else
>> >>> - /* no valid new
>> >>> master */
>> >>> -
>> >>> string_append_char(exec_cmd, "");
>> >>> - break;
>> >>> -
>> >>> - case 'M': /* old master node
>> id */
>> >>> - snprintf(port_buf,
>> >>> sizeof(port_buf), "%d", old_master);
>> >>> -
>> >>> string_append_char(exec_cmd, port_buf);
>> >>> - break;
>> >>> -
>> >>> - case 'P': /* old primary node
>> id
>> >>> */
>> >>> - snprintf(port_buf,
>> >>> sizeof(port_buf), "%d", old_primary);
>> >>> -
>> >>> string_append_char(exec_cmd, port_buf);
>> >>> - break;
>> >>> -
>> >>> - case '%': /* escape */
>> >>> -
>> >>> string_append_char(exec_cmd, "%");
>> >>> - break;
>> >>> -
>> >>> - default: /* ignore */
>> >>> - break;
>> >>> - }
>> >>> - command_line++;
>> >>> - }
>> >>> - } else {
>> >>> - buf[0] = *command_line;
>> >>> - string_append_char(exec_cmd, buf);
>> >>> - }
>> >>> - command_line++;
>> >>> - }
>> >>> -
>> >>> - if (strlen(exec_cmd->data) != 0)
>> >>> - {
>> >>> - pool_log("execute command: %s", exec_cmd->data);
>> >>> - r = system(exec_cmd->data);
>> >>> - }
>> >>> -
>> >>> - pool_memory_delete(pool_memory, 0);
>> >>> - pool_memory = NULL;
>> >>> -
>> >>> - return r;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> - * Find the primary node (i.e. not standby node) and returns its node
>> >>> - * id. If no primary node is found, returns -1.
>> >>> - */
>> >>> -static int find_primary_node(void)
>> >>> -{
>> >>> - BackendInfo *bkinfo;
>> >>> - POOL_CONNECTION_POOL_SLOT *s;
>> >>> - POOL_CONNECTION *con;
>> >>> - POOL_STATUS status;
>> >>> - POOL_SELECT_RESULT *res;
>> >>> - bool is_standby;
>> >>> - int i;
>> >>> -
>> >>> - /* Streaming replication mode? */
>> >>> - if (pool_config->master_slave_mode == 0 ||
>> >>> - strcmp(pool_config->master_slave_sub_mode,
>> >>> MODE_STREAMREP))
>> >>> - {
>> >>> - /* No point to look for primary node if not in
>> streaming
>> >>> - * replication mode.
>> >>> - */
>> >>> - pool_debug("find_primary_node: not in streaming
>> >>> replication mode");
>> >>> - return -1;
>> >>> - }
>> >>> -
>> >>> - for(i=0;i<NUM_BACKENDS;i++)
>> >>> - {
>> >>> - if (!VALID_BACKEND(i))
>> >>> - continue;
>> >>> -
>> >>> - /*
>> >>> - * Check to see if this is a standby node or not.
>> >>> - */
>> >>> - is_standby = false;
>> >>> -
>> >>> - bkinfo = pool_get_node_info(i);
>> >>> - s =
>> >>> make_persistent_db_connection(bkinfo->backend_hostname,
>> >>> -
>> >>> bkinfo->backend_port,
>> >>> -
>> >>> "postgres",
>> >>> -
>> >>> pool_config->sr_check_user,
>> >>> -
>> >>> pool_config->sr_check_password, true);
>> >>> - if (!s)
>> >>> - {
>> >>> - pool_error("find_primary_node:
>> >>> make_persistent_connection failed");
>> >>> - return -1;
>> >>> - }
>> >>> - con = s->con;
>> >>> - status = do_query(con, "SELECT pg_is_in_recovery()",
>> >>> - &res,
>> PROTO_MAJOR_V3);
>> >>> - if (res->numrows <= 0)
>> >>> - {
>> >>> - pool_log("find_primary_node: do_query returns
>> no
>> >>> rows");
>> >>> - }
>> >>> - if (res->data[0] == NULL)
>> >>> - {
>> >>> - pool_log("find_primary_node: do_query returns
>> no
>> >>> data");
>> >>> - }
>> >>> - if (res->nullflags[0] == -1)
>> >>> - {
>> >>> - pool_log("find_primary_node: do_query returns
>> >>> NULL");
>> >>> - }
>> >>> - if (res->data[0] && !strcmp(res->data[0], "t"))
>> >>> - {
>> >>> - is_standby = true;
>> >>> - }
>> >>> - free_select_result(res);
>> >>> - discard_persistent_db_connection(s);
>> >>> -
>> >>> - /*
>> >>> - * If this is a standby, we continue to look for
>> primary
>> >>> node.
>> >>> - */
>> >>> - if (is_standby)
>> >>> - {
>> >>> - pool_debug("find_primary_node: %d node is
>> >>> standby", i);
>> >>> - }
>> >>> - else
>> >>> - {
>> >>> - break;
>> >>> - }
>> >>> - }
>> >>> -
>> >>> - if (i == NUM_BACKENDS)
>> >>> - {
>> >>> - pool_debug("find_primary_node: no primary node found");
>> >>> - return -1;
>> >>> - }
>> >>> -
>> >>> - pool_log("find_primary_node: primary node id is %d", i);
>> >>> - return i;
>> >>> -}
>> >>> -
>> >>> -static int find_primary_node_repeatedly(void)
>> >>> +/* Call back function to unlink the file */
>> >>> +/* Call back function to unlink the file */
>> >>> +static void FileUnlink(int code, Datum path)
>> >>> {
>> >>> - int sec;
>> >>> - int node_id = -1;
>> >>> -
>> >>> - /* Streaming replication mode? */
>> >>> - if (pool_config->master_slave_mode == 0 ||
>> >>> - strcmp(pool_config->master_slave_sub_mode,
>> >>> MODE_STREAMREP))
>> >>> - {
>> >>> - /* No point to look for primary node if not in
>> streaming
>> >>> - * replication mode.
>> >>> - */
>> >>> - pool_debug("find_primary_node: not in streaming
>> >>> replication mode");
>> >>> - return -1;
>> >>> - }
>> >>> -
>> >>> - /*
>> >>> - * Try to find the new primary node and keep trying for
>> >>> - * search_primary_node_timeout seconds.
>> >>> - * search_primary_node_timeout = 0 means never timeout and keep
>> >>> searching
>> >>> - * indefinitely
>> >>> + char* filePath = (char*)path;
>> >>> + if (unlink(filePath) == 0) return;
>> >>> + /*
>> >>> + * We are already exiting the system just produce a log entry
>> to
>> >>> report an error
>> >>> */
>> >>> - pool_log("find_primary_node_repeatedly: waiting for finding a
>> >>> primary node");
>> >>> - for (sec = 0; (pool_config->search_primary_node_timeout == 0 ||
>> >>> - sec <
>> >>> pool_config->search_primary_node_timeout); sec++)
>> >>> - {
>> >>> - node_id = find_primary_node();
>> >>> - if (node_id != -1)
>> >>> - break;
>> >>> - pool_sleep(1);
>> >>> - }
>> >>> - return node_id;
>> >>> -}
>> >>> -
>> >>> -/*
>> >>> -* fork a follow child
>> >>> -*/
>> >>> -pid_t fork_follow_child(int old_master, int new_primary, int
>> old_primary)
>> >>> -{
>> >>> - pid_t pid;
>> >>> - int i;
>> >>> -
>> >>> - pid = fork();
>> >>> -
>> >>> - if (pid == 0)
>> >>> - {
>> >>> - pool_log("start triggering follow command.");
>> >>> - for (i = 0; i <
>> pool_config->backend_desc->num_backends;
>> >>> i++)
>> >>> - {
>> >>> - BackendInfo *bkinfo;
>> >>> - bkinfo = pool_get_node_info(i);
>> >>> - if (bkinfo->backend_status == CON_DOWN)
>> >>> - trigger_failover_command(i,
>> >>> pool_config->follow_master_command,
>> >>> -
>> >>> old_master, new_primary, old_primary);
>> >>> - }
>> >>> - exit(0);
>> >>> - }
>> >>> - else if (pid == -1)
>> >>> - {
>> >>> - pool_error("follow fork() failed. reason: %s",
>> >>> strerror(errno));
>> >>> - exit(1);
>> >>> - }
>> >>> - return pid;
>> >>> + ereport(LOG,
>> >>> + (errmsg("unlink failed for file at path \"%s\"",
>> >>> filePath),
>> >>> + errdetail("\"%s\"", strerror(errno))));
>> >>> }
>> >>>
>> >>> ...
>> >>>
>> >>> [Message clipped]
>> >>
>> >>
>> >>
>> >>
>> >> --
>> >> Ahsan Hadi
>> >> Snr Director Product Development
>> >> EnterpriseDB Corporation
>> >> The Enterprise Postgres Company
>> >>
>> >> Phone: +92-51-8358874
>> >> Mobile: +92-333-5162114
>> >>
>> >> Website: www.enterprisedb.com
>> >> EnterpriseDB Blog: http://blogs.enterprisedb.com/
>> >> Follow us on Twitter: http://www.twitter.com/enterprisedb
>> >>
>> >> This e-mail message (and any attachment) is intended for the use of the
>> >> individual or entity to whom it is addressed. This message contains
>> >> information from EnterpriseDB Corporation that may be privileged,
>> >> confidential, or exempt from disclosure under applicable law. If you are
>> >> not the intended recipient or authorized to receive this for the
>> intended
>> >> recipient, any use, dissemination, distribution, retention, archiving,
>> or
>> >> copying of this communication is strictly prohibited. If you have
>> received
>> >> this e-mail in error, please notify the sender immediately by reply
>> e-mail
>> >> and delete this message.
>> >>
>>
More information about the pgpool-hackers
mailing list