big-data/hadoop
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

YARN-4459. container-executor should only kill process groups. Contributed by Jun Gong

Browse Source
This commit is contained in:
Jason Lowe 2016-05-25 21:24:55 +00:00
parent f69f5ab3b6
commit 1ba31fe9e9
2 changed files with 5 additions and 63 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/impl/container-executor.c
View File

@ -1540,21 +1540,13 @@ int signal_container_as_user(const char *user, int pid, int sig) {
} }
//Don't continue if the process-group is not alive anymore. //Don't continue if the process-group is not alive anymore.
int has_group = 1;
if (kill(-pid,0) < 0) { if (kill(-pid,0) < 0) {
if (kill(pid, 0) < 0) { fprintf(LOGFILE, "Error signalling not exist process group %d "
if (errno == ESRCH) { "with signal %d\n", pid, sig);
return INVALID_CONTAINER_PID; return INVALID_CONTAINER_PID;
} }
fprintf(LOGFILE, "Error signalling container %d with %d - %s\n",
pid, sig, strerror(errno));
return -1;
} else {
has_group = 0;
}
}
if (kill((has_group ? -1 : 1) * pid, sig) < 0) { if (kill(-pid, sig) < 0) {
if(errno != ESRCH) { if(errno != ESRCH) {
fprintf(LOGFILE, fprintf(LOGFILE,
"Error signalling process group %d with signal %d - %s\n", "Error signalling process group %d with signal %d - %s\n",
@ -1568,8 +1560,7 @@ int signal_container_as_user(const char *user, int pid, int sig) {
return INVALID_CONTAINER_PID; return INVALID_CONTAINER_PID;
} }
} }
fprintf(LOGFILE, "Killing process %s%d with %d\n", fprintf(LOGFILE, "Killing process group %d with %d\n", pid, sig);
(has_group ? "group " :""), pid, sig);
return 0; return 0;
} }

49
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/test/test-container-executor.c
View File

@ -471,54 +471,6 @@ void run_test_in_child(const char* test_name, void (*func)()) {
} }
} }
void test_signal_container() {
sigset_t set;
// unblock SIGQUIT
sigemptyset(&set);
sigaddset(&set, SIGQUIT);
sigprocmask(SIG_UNBLOCK, &set, NULL);
printf("\nTesting signal_container\n");
fflush(stdout);
fflush(stderr);
pid_t child = fork();
if (child == -1) {
printf("FAIL: fork failed\n");
exit(1);
} else if (child == 0) {
printf("\nSwitching to user %d\n", user_detail->pw_uid);
if (change_user(user_detail->pw_uid, user_detail->pw_gid) != 0) {
exit(1);
}
sleep(3600);
exit(0);
} else {
printf("Child container launched as %" PRId64 "\n", (int64_t)child);
printf("Signaling container as user %s\n", yarn_username);
// there's a race condition for child calling change_user and us
// calling signal_container_as_user, hence sleeping
sleep(3);
if (signal_container_as_user(yarn_username, child, SIGQUIT) != 0) {
exit(1);
}
int status = 0;
if (waitpid(child, &status, 0) == -1) {
printf("FAIL: waitpid failed - %s\n", strerror(errno));
exit(1);
}
if (!WIFSIGNALED(status)) {
printf("FAIL: child wasn't signalled - %d\n", status);
exit(1);
}
if (WTERMSIG(status) != SIGQUIT) {
printf("FAIL: child was killed with %d instead of %d\n",
WTERMSIG(status), SIGQUIT);
exit(1);
}
}
}
void test_signal_container_group() { void test_signal_container_group() {
printf("\nTesting group signal_container\n"); printf("\nTesting group signal_container\n");
fflush(stdout); fflush(stdout);
@ -931,7 +883,6 @@ int main(int argc, char **argv) {
// the tests that change user need to be run in a subshell, so that // the tests that change user need to be run in a subshell, so that
// when they change user they don't give up our privs // when they change user they don't give up our privs
run_test_in_child("test_signal_container", test_signal_container);
run_test_in_child("test_signal_container_group", test_signal_container_group); run_test_in_child("test_signal_container_group", test_signal_container_group);
// init app and run container can't be run if you aren't testing as root // init app and run container can't be run if you aren't testing as root