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

Commit99351a8

Browse files
committed
Improve isolationtester's timeout management.
isolationtester.c had a hard-wired limit of 3 minutes per test step.It now emerges that this isn't quite enough for some of the slowestbuildfarm animals. This isn't the first time we've had to raisethis limit (cf.1db439a), so let's make it configurable. Thispatch raises the default to 5 minutes, and introduces an environmentvariable PGISOLATIONTIMEOUT that can be set if more time is needed,following the precedent of PGCTLTIMEOUT.Also, modify isolationtester so that when the timeout is hit,it explicitly reports having sent a cancel. This makes the regressionfailure log considerably more intelligible. (In the worst case, atimed-out test might actually be reported as "passing" without thisextra output, so arguably this is a bug fix in itself.)In passing, update the README file, which had apparently not gottentouched when we added "make check" support here.Back-patch to 9.6; older versions don't have comparable timeout logic.Discussion:https://postgr.es/m/22964.1575842935@sss.pgh.pa.us
1 parent2d0fdfa commit99351a8

File tree

2 files changed

+48
-14
lines changed

2 files changed

+48
-14
lines changed

‎src/test/isolation/README

Lines changed: 21 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -11,18 +11,26 @@ comes from the fact that the original motivation was to test the
1111
serializable isolation level; but tests for other sorts of concurrent
1212
behaviors have been added as well.
1313

14-
Torun the tests, you need to have a server running at the default port
15-
expected by libpq. (You can set PGPORT and so forth in your environment
16-
to control this.) Then run
14+
You canrun the tests against the current build tree by typing
15+
make check
16+
Alternatively, you can run against an existing installation by typing
1717
make installcheck
18-
To run just specific test(s), you can do something like
18+
(This will contact a server at the default port expected by libpq.
19+
You can set PGPORT and so forth in your environment to control this.)
20+
21+
To run just specific test(s) against an installed server,
22+
you can do something like
1923
./pg_isolation_regress fk-contention fk-deadlock
2024
(look into the specs/ subdirectory to see the available tests).
2125

2226
The prepared-transactions test requires the server's
2327
max_prepared_transactions parameter to be set to at least 3; therefore it
2428
is not run by default. To include it in the test run, use
29+
make check-prepared-txns
30+
or
2531
make installcheck-prepared-txns
32+
after making sure the server configuration is correct (see TEMP_CONFIG
33+
to adjust this in the "check" case).
2634

2735
To define tests with overlapping transactions, we use test specification
2836
files with a custom syntax, which is described in the next section. To add
@@ -39,6 +47,11 @@ pg_isolation_regress is a tool similar to pg_regress, but instead of using
3947
psql to execute a test, it uses isolationtester. It accepts all the same
4048
command-line arguments as pg_regress.
4149

50+
By default, isolationtester will wait at most 300 seconds (5 minutes)
51+
for any one test step to complete. If you need to adjust this, set
52+
the environment variable PGISOLATIONTIMEOUT to the desired timeout
53+
in seconds.
54+
4255

4356
Test specification
4457
==================
@@ -108,10 +121,10 @@ Each step may contain commands that block until further action has been taken
108121
deadlock). A test that uses this ability must manually specify valid
109122
permutations, i.e. those that would not expect a blocked session to execute a
110123
command. If a test fails to follow that rule, isolationtester will cancel it
111-
after180 seconds. If the cancel doesn't work, isolationtester will exit
112-
uncleanly after a total of200 seconds of wait time. Testing invalid
113-
permutations should be avoided because they can make the isolation tests take
114-
a very long time to run, and they serve no useful testing purpose.
124+
afterPGISOLATIONTIMEOUT seconds. If the cancel doesn't work, isolationtester
125+
will exituncleanly after a total oftwice PGISOLATIONTIMEOUT. Testing
126+
invalidpermutations should be avoided because they can make the isolation
127+
tests takea very long time to run, and they serve no useful testing purpose.
115128

116129
Note that isolationtester recognizes that a command has blocked by looking
117130
to see if it is shown as waiting in the pg_locks view; therefore, only

‎src/test/isolation/isolationtester.c

Lines changed: 27 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -30,6 +30,10 @@ static int *backend_pids = NULL;
3030
staticconstchar**backend_pid_strs=NULL;
3131
staticintnconns=0;
3232

33+
/* Maximum time to wait before giving up on a step (in usec) */
34+
staticint64max_step_wait=300*USECS_PER_SEC;
35+
36+
3337
staticvoidrun_testspec(TestSpec*testspec);
3438
staticvoidrun_all_permutations(TestSpec*testspec);
3539
staticvoidrun_all_permutations_recurse(TestSpec*testspec,intnsteps,
@@ -62,6 +66,7 @@ int
6266
main(intargc,char**argv)
6367
{
6468
constchar*conninfo;
69+
constchar*env_wait;
6570
TestSpec*testspec;
6671
inti,
6772
j;
@@ -103,6 +108,14 @@ main(int argc, char **argv)
103108
else
104109
conninfo="dbname = postgres";
105110

111+
/*
112+
* If PGISOLATIONTIMEOUT is set in the environment, adopt its value (given
113+
* in seconds) as the max time to wait for any one step to complete.
114+
*/
115+
env_wait=getenv("PGISOLATIONTIMEOUT");
116+
if (env_wait!=NULL)
117+
max_step_wait= ((int64)atoi(env_wait))*USECS_PER_SEC;
118+
106119
/* Read the test spec from stdin */
107120
spec_yyparse();
108121
testspec=&parseresult;
@@ -766,15 +779,15 @@ try_complete_step(TestSpec *testspec, Step *step, int flags)
766779
td+= (int64)current_time.tv_usec- (int64)start_time.tv_usec;
767780

768781
/*
769-
* After180 seconds, try to cancel the query.
782+
* Aftermax_step_wait microseconds, try to cancel the query.
770783
*
771784
* If the user tries to test an invalid permutation, we don't want
772785
* to hang forever, especially when this is running in the
773786
* buildfarm. This will presumably lead to this permutation
774787
* failing, but remaining permutations and tests should still be
775788
* OK.
776789
*/
777-
if (td>180*USECS_PER_SEC&& !canceled)
790+
if (td>max_step_wait&& !canceled)
778791
{
779792
PGcancel*cancel=PQgetCancel(conn);
780793

@@ -783,24 +796,32 @@ try_complete_step(TestSpec *testspec, Step *step, int flags)
783796
charbuf[256];
784797

785798
if (PQcancel(cancel,buf,sizeof(buf)))
799+
{
800+
/*
801+
* print to stdout not stderr, as this should appear
802+
* in the test case's results
803+
*/
804+
printf("isolationtester: canceling step %s after %d seconds\n",
805+
step->name, (int) (td /USECS_PER_SEC));
786806
canceled= true;
807+
}
787808
else
788809
fprintf(stderr,"PQcancel failed: %s\n",buf);
789810
PQfreeCancel(cancel);
790811
}
791812
}
792813

793814
/*
794-
* After200 seconds, just give up and die.
815+
* Aftertwice max_step_wait, just give up and die.
795816
*
796817
* Since cleanup steps won't be run in this case, this may cause
797818
* later tests to fail. That stinks, but it's better than waiting
798819
* forever for the server to respond to the cancel.
799820
*/
800-
if (td>200*USECS_PER_SEC)
821+
if (td>2*max_step_wait)
801822
{
802-
fprintf(stderr,"step %s timed out after200 seconds\n",
803-
step->name);
823+
fprintf(stderr,"step %s timed out after%d seconds\n",
824+
step->name, (int) (td /USECS_PER_SEC));
804825
exit(1);
805826
}
806827
}

0 commit comments

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