Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit733cf37

Browse files
author
Artur Zakirov
committed
Use pg_probackup.pid file instead of locking pg_probackup.conf
1 parentb67dd1e commit733cf37

File tree

7 files changed

+178
-87
lines changed

7 files changed

+178
-87
lines changed

‎backup.c

Lines changed: 1 addition & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -404,7 +404,6 @@ do_backup_database(parray *backup_list, bool smooth_checkpoint)
404404
int
405405
do_backup(boolsmooth_checkpoint)
406406
{
407-
intret;
408407
parray*backup_list;
409408
parray*files_database;
410409

@@ -433,12 +432,7 @@ do_backup(bool smooth_checkpoint)
433432
elog(LOG,"----------------------------------------");
434433

435434
/* get exclusive lock of backup catalog */
436-
ret=catalog_lock(true);
437-
if (ret==-1)
438-
elog(ERROR,"cannot lock backup catalog");
439-
elseif (ret==1)
440-
elog(ERROR,
441-
"another pg_probackup is running, skipping this backup");
435+
catalog_lock(true);
442436

443437
/* initialize backup result */
444438
current.status=BACKUP_STATUS_RUNNING;
@@ -509,9 +503,6 @@ do_backup(bool smooth_checkpoint)
509503

510504
pgBackupValidate(&current, false, false);
511505

512-
/* release catalog lock */
513-
catalog_unlock();
514-
515506
return0;
516507
}
517508

‎catalog.c

Lines changed: 167 additions & 33 deletions
Original file line numberDiff line numberDiff line change
@@ -13,6 +13,7 @@
1313
#include<dirent.h>
1414
#include<fcntl.h>
1515
#include<libgen.h>
16+
#include<signal.h>
1617
#include<sys/file.h>
1718
#include<sys/stat.h>
1819
#include<sys/types.h>
@@ -26,42 +27,185 @@ static pgBackup *read_backup_from_file(const char *path);
2627

2728
#defineBOOL_TO_STR(val)((val) ? "true" : "false")
2829

29-
staticintlock_fd=-1;
30+
staticboolexit_hook_registered= false;
31+
staticcharlock_file[MAXPGPATH];
32+
33+
staticvoid
34+
unlink_lock_atexit(void)
35+
{
36+
intres;
37+
res=unlink(lock_file);
38+
if (res!=0&&res!=ENOENT)
39+
elog(WARNING,"%s: %s",lock_file,strerror(errno));
40+
}
3041

3142
/*
32-
* Lock of the catalog with pg_probackup.conf file and return 0.
33-
* If the lock is held by another one, return 1 immediately.
43+
* Create a lockfile.
3444
*/
3545
int
3646
catalog_lock(boolcheck_catalog)
3747
{
38-
intret;
39-
charid_path[MAXPGPATH];
40-
41-
join_path_components(id_path,backup_path,BACKUP_CATALOG_CONF_FILE);
42-
lock_fd=open(id_path,O_RDWR);
43-
if (lock_fd==-1)
44-
elog(errno==ENOENT ?ERROR :ERROR,
45-
"cannot open file \"%s\": %s",id_path,strerror(errno));
46-
#ifdef__IBMC__
47-
ret=lockf(lock_fd,LOCK_EX |LOCK_NB,0);/* non-blocking */
48+
intfd;
49+
charbuffer[MAXPGPATH*2+256];
50+
intntries;
51+
intlen;
52+
intencoded_pid;
53+
pid_tmy_pid,
54+
my_p_pid;
55+
56+
join_path_components(lock_file,backup_path,BACKUP_CATALOG_PID);
57+
58+
/*
59+
* If the PID in the lockfile is our own PID or our parent's or
60+
* grandparent's PID, then the file must be stale (probably left over from
61+
* a previous system boot cycle). We need to check this because of the
62+
* likelihood that a reboot will assign exactly the same PID as we had in
63+
* the previous reboot, or one that's only one or two counts larger and
64+
* hence the lockfile's PID now refers to an ancestor shell process. We
65+
* allow pg_ctl to pass down its parent shell PID (our grandparent PID)
66+
* via the environment variable PG_GRANDPARENT_PID; this is so that
67+
* launching the postmaster via pg_ctl can be just as reliable as
68+
* launching it directly. There is no provision for detecting
69+
* further-removed ancestor processes, but if the init script is written
70+
* carefully then all but the immediate parent shell will be root-owned
71+
* processes and so the kill test will fail with EPERM. Note that we
72+
* cannot get a false negative this way, because an existing postmaster
73+
* would surely never launch a competing postmaster or pg_ctl process
74+
* directly.
75+
*/
76+
my_pid=getpid();
77+
#ifndefWIN32
78+
my_p_pid=getppid();
4879
#else
49-
ret=flock(lock_fd,LOCK_EX |LOCK_NB);/* non-blocking */
80+
81+
/*
82+
* Windows hasn't got getppid(), but doesn't need it since it's not using
83+
* real kill() either...
84+
*/
85+
my_p_pid=0;
5086
#endif
51-
if (ret==-1)
87+
88+
/*
89+
* We need a loop here because of race conditions. But don't loop forever
90+
* (for example, a non-writable $backup_path directory might cause a failure
91+
* that won't go away). 100 tries seems like plenty.
92+
*/
93+
for (ntries=0;;ntries++)
5294
{
53-
if (errno==EWOULDBLOCK)
95+
/*
96+
* Try to create the lock file --- O_EXCL makes this atomic.
97+
*
98+
* Think not to make the file protection weaker than 0600. See
99+
* comments below.
100+
*/
101+
fd=open(lock_file,O_RDWR |O_CREAT |O_EXCL,0600);
102+
if (fd >=0)
103+
break;/* Success; exit the retry loop */
104+
105+
/*
106+
* Couldn't create the pid file. Probably it already exists.
107+
*/
108+
if ((errno!=EEXIST&&errno!=EACCES)||ntries>100)
109+
elog(ERROR,"could not create lock file \"%s\": %s",
110+
lock_file,strerror(errno));
111+
112+
/*
113+
* Read the file to get the old owner's PID. Note race condition
114+
* here: file might have been deleted since we tried to create it.
115+
*/
116+
fd=open(lock_file,O_RDONLY,0600);
117+
if (fd<0)
54118
{
55-
close(lock_fd);
56-
return1;
119+
if (errno==ENOENT)
120+
continue;/* race condition; try again */
121+
elog(ERROR,"could not open lock file \"%s\": %s",
122+
lock_file,strerror(errno));
57123
}
58-
else
124+
if ((len=read(fd,buffer,sizeof(buffer)-1))<0)
125+
elog(ERROR,"could not read lock file \"%s\": %s",
126+
lock_file,strerror(errno));
127+
close(fd);
128+
129+
if (len==0)
130+
elog(ERROR,"lock file \"%s\" is empty",lock_file);
131+
132+
buffer[len]='\0';
133+
encoded_pid=atoi(buffer);
134+
135+
if (encoded_pid <=0)
136+
elog(ERROR,"bogus data in lock file \"%s\": \"%s\"",
137+
lock_file,buffer);
138+
139+
/*
140+
* Check to see if the other process still exists
141+
*
142+
* Per discussion above, my_pid, my_p_pid can be
143+
* ignored as false matches.
144+
*
145+
* Normally kill() will fail with ESRCH if the given PID doesn't
146+
* exist.
147+
*/
148+
if (encoded_pid!=my_pid&&encoded_pid!=my_p_pid)
59149
{
60-
interrno_tmp=errno;
61-
close(lock_fd);
62-
elog(ERROR,"cannot lock file \"%s\": %s",id_path,
63-
strerror(errno_tmp));
150+
if (kill(encoded_pid,0)==0||
151+
(errno!=ESRCH&&errno!=EPERM))
152+
elog(ERROR,"lock file \"%s\" already exists",lock_file);
64153
}
154+
155+
/*
156+
* Looks like nobody's home. Unlink the file and try again to create
157+
* it. Need a loop because of possible race condition against other
158+
* would-be creators.
159+
*/
160+
if (unlink(lock_file)<0)
161+
elog(ERROR,"could not remove old lock file \"%s\": %s",
162+
lock_file,strerror(errno));
163+
}
164+
165+
/*
166+
* Successfully created the file, now fill it.
167+
*/
168+
snprintf(buffer,sizeof(buffer),"%d\n",my_pid);
169+
170+
errno=0;
171+
if (write(fd,buffer,strlen(buffer))!=strlen(buffer))
172+
{
173+
intsave_errno=errno;
174+
175+
close(fd);
176+
unlink(lock_file);
177+
/* if write didn't set errno, assume problem is no disk space */
178+
errno=save_errno ?save_errno :ENOSPC;
179+
elog(ERROR,"could not write lock file \"%s\": %s",
180+
lock_file,strerror(errno));
181+
}
182+
if (fsync(fd)!=0)
183+
{
184+
intsave_errno=errno;
185+
186+
close(fd);
187+
unlink(lock_file);
188+
errno=save_errno;
189+
elog(ERROR,"could not write lock file \"%s\": %s",
190+
lock_file,strerror(errno));
191+
}
192+
if (close(fd)!=0)
193+
{
194+
intsave_errno=errno;
195+
196+
unlink(lock_file);
197+
errno=save_errno;
198+
elog(ERROR,"could not write lock file \"%s\": %s",
199+
lock_file,strerror(errno));
200+
}
201+
202+
/*
203+
* Arrange to unlink the lock file(s) at proc_exit.
204+
*/
205+
if (!exit_hook_registered)
206+
{
207+
atexit(unlink_lock_atexit);
208+
exit_hook_registered= true;
65209
}
66210

67211
if (check_catalog)
@@ -80,16 +224,6 @@ catalog_lock(bool check_catalog)
80224
return0;
81225
}
82226

83-
/*
84-
* Release catalog lock.
85-
*/
86-
void
87-
catalog_unlock(void)
88-
{
89-
close(lock_fd);
90-
lock_fd=-1;
91-
}
92-
93227
/*
94228
* Create a pgBackup which taken at timestamp.
95229
* If no backup matches, return NULL.

‎delete.c

Lines changed: 6 additions & 28 deletions
Original file line numberDiff line numberDiff line change
@@ -23,7 +23,6 @@ do_delete(time_t backup_id)
2323
{
2424
inti;
2525
intb_index;
26-
intret;
2726
parray*backup_list;
2827
pgBackup*last_backup=NULL;
2928

@@ -32,12 +31,7 @@ do_delete(time_t backup_id)
3231
elog(ERROR,"required backup ID not specified");
3332

3433
/* Lock backup catalog */
35-
ret=catalog_lock(false);
36-
if (ret==-1)
37-
elog(ERROR,"can't lock backup catalog.");
38-
elseif (ret==1)
39-
elog(ERROR,
40-
"another pg_probackup is running, stop delete.");
34+
catalog_lock(false);
4135

4236
/* Get complete list of backups */
4337
backup_list=catalog_get_backup_list(0);
@@ -78,15 +72,12 @@ do_delete(time_t backup_id)
7872
pgBackupDeleteFiles(backup);
7973
}
8074

81-
/* release catalog lock */
82-
catalog_unlock();
83-
8475
/* cleanup */
8576
parray_walk(backup_list,pgBackupFree);
8677
parray_free(backup_list);
8778

8879
if (delete_wal)
89-
do_deletewal(backup_id, false);
80+
do_deletewal(backup_id, false, false);
9081

9182
return0;
9283
}
@@ -97,22 +88,17 @@ do_delete(time_t backup_id)
9788
* found around needs to keep.
9889
*/
9990
int
100-
do_deletewal(time_tbackup_id,boolstrict)
91+
do_deletewal(time_tbackup_id,boolstrict,boolneed_catalog_lock)
10192
{
10293
size_ti;
103-
intret;
10494
parray*backup_list;
10595
XLogRecPtroldest_lsn=InvalidXLogRecPtr;
10696
TimeLineIDoldest_tli;
10797
boolbackup_found= false;
10898

10999
/* Lock backup catalog */
110-
ret=catalog_lock(false);
111-
if (ret==-1)
112-
elog(ERROR,"can't lock backup catalog.");
113-
elseif (ret==1)
114-
elog(ERROR,
115-
"another pg_probackup is running, stop delete.");
100+
if (need_catalog_lock)
101+
catalog_lock(false);
116102

117103
/* Find oldest LSN, used by backups */
118104
backup_list=catalog_get_backup_list(0);
@@ -136,7 +122,6 @@ do_deletewal(time_t backup_id, bool strict)
136122
if (strict&&backup_id!=0&&backup_found== false)
137123
elog(ERROR,"not found backup for deletwal command");
138124

139-
catalog_unlock();
140125
parray_walk(backup_list,pgBackupFree);
141126
parray_free(backup_list);
142127

@@ -158,7 +143,6 @@ do_retention_purge(void)
158143
time_tdays_threshold=time(NULL)- (retention_window*60*60*24);
159144
XLogRecPtroldest_lsn=InvalidXLogRecPtr;
160145
TimeLineIDoldest_tli;
161-
intret;
162146
boolkeep_next_backup= true;/* Do not delete first full backup */
163147

164148
if (retention_redundancy>0)
@@ -170,18 +154,14 @@ do_retention_purge(void)
170154
elog(ERROR,"retention policy is not set");
171155

172156
/* Lock backup catalog */
173-
ret=catalog_lock(false);
174-
if (ret==1)
175-
elog(ERROR,
176-
"cannot lock backup catalog, another pg_probackup is running");
157+
catalog_lock(false);
177158

178159
/* Get a complete list of backups. */
179160
backup_list=catalog_get_backup_list(0);
180161
if (parray_num(backup_list)==0)
181162
{
182163
elog(INFO,"backup list is empty");
183164
elog(INFO,"exit");
184-
catalog_unlock();
185165
return0;
186166
}
187167

@@ -236,8 +216,6 @@ do_retention_purge(void)
236216
parray_walk(backup_list,pgBackupFree);
237217
parray_free(backup_list);
238218

239-
catalog_unlock();
240-
241219
elog(INFO,"purging is finished");
242220

243221
return0;

‎pg_probackup.c

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -209,7 +209,7 @@ main(int argc, char *argv[])
209209
elseif (pg_strcasecmp(cmd,"delete")==0)
210210
returndo_delete(backup_id);
211211
elseif (pg_strcasecmp(cmd,"delwal")==0)
212-
returndo_deletewal(backup_id, true);
212+
returndo_deletewal(backup_id, true, true);
213213
elseif (pg_strcasecmp(cmd,"retention")==0)
214214
{
215215
if (subcmd==NULL)

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp