Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

YARN-7189. Container-executor doesn't remove Docker containers that error out early. Contributed by Eric Badger 

(cherry picked from commit 391ac5cdd2)
This commit is contained in:
Jason Lowe 2018-04-17 09:45:55 -05:00
parent 88cb461c87
commit 5ec195edbc
1 changed files with 46 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

63
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/impl/container-executor.c
View File

@ -1444,7 +1444,7 @@ int launch_docker_container_as_user(const char * user, const char *app_id,
if (exit_code != 0) { if (exit_code != 0) {
fprintf(ERRORFILE, "Could not create script path\n"); fprintf(ERRORFILE, "Could not create script path\n");
fflush(ERRORFILE); fflush(ERRORFILE);
goto cleanup; goto pre_launch_cleanup;
} }
fprintf(LOGFILE, "Creating local dirs...\n"); fprintf(LOGFILE, "Creating local dirs...\n");
@ -1455,7 +1455,7 @@ int launch_docker_container_as_user(const char * user, const char *app_id,
if (exit_code != 0) { if (exit_code != 0) {
fprintf(ERRORFILE, "Could not create local files and directories %d %d\n", container_file_source, cred_file_source); fprintf(ERRORFILE, "Could not create local files and directories %d %d\n", container_file_source, cred_file_source);
fflush(ERRORFILE); fflush(ERRORFILE);
goto cleanup; goto pre_launch_cleanup;
} }
docker_command = construct_docker_command(command_file); docker_command = construct_docker_command(command_file);
@ -1467,14 +1467,14 @@ int launch_docker_container_as_user(const char * user, const char *app_id,
exit_code = OUT_OF_MEMORY; exit_code = OUT_OF_MEMORY;
fprintf(ERRORFILE, "Container out of memory"); fprintf(ERRORFILE, "Container out of memory");
fflush(ERRORFILE); fflush(ERRORFILE);
goto cleanup; goto pre_launch_cleanup;
} }
fprintf(LOGFILE, "Changing effective user to root...\n"); fprintf(LOGFILE, "Changing effective user to root...\n");
if (change_effective_user(0, user_gid) != 0) { if (change_effective_user(0, user_gid) != 0) {
fprintf(ERRORFILE, "Could not change to effective users %d, %d\n", 0, user_gid); fprintf(ERRORFILE, "Could not change to effective users %d, %d\n", 0, user_gid);
fflush(ERRORFILE); fflush(ERRORFILE);
goto cleanup; goto pre_launch_cleanup;
} }
snprintf(docker_command_with_binary, command_size, "%s %s", docker_binary, docker_command); snprintf(docker_command_with_binary, command_size, "%s %s", docker_binary, docker_command);
@ -1487,7 +1487,7 @@ int launch_docker_container_as_user(const char * user, const char *app_id,
"Could not invoke docker %s.\n", docker_command_with_binary); "Could not invoke docker %s.\n", docker_command_with_binary);
fflush(ERRORFILE); fflush(ERRORFILE);
exit_code = UNABLE_TO_EXECUTE_CONTAINER_SCRIPT; exit_code = UNABLE_TO_EXECUTE_CONTAINER_SCRIPT;
goto cleanup; goto post_launch_cleanup;
} }
snprintf(docker_inspect_command, command_size, snprintf(docker_inspect_command, command_size,
@ -1504,7 +1504,7 @@ int launch_docker_container_as_user(const char * user, const char *app_id,
"Could not inspect docker to get pid %s.\n", docker_inspect_command); "Could not inspect docker to get pid %s.\n", docker_inspect_command);
fflush(ERRORFILE); fflush(ERRORFILE);
exit_code = UNABLE_TO_EXECUTE_CONTAINER_SCRIPT; exit_code = UNABLE_TO_EXECUTE_CONTAINER_SCRIPT;
goto cleanup; goto post_launch_cleanup;
} }
if (pid != 0) { if (pid != 0) {
@ -1519,7 +1519,7 @@ int launch_docker_container_as_user(const char * user, const char *app_id,
if (strcmp(*cgroup_ptr, "none") != 0 && if (strcmp(*cgroup_ptr, "none") != 0 &&
write_pid_to_cgroup_as_root(*cgroup_ptr, pid) != 0) { write_pid_to_cgroup_as_root(*cgroup_ptr, pid) != 0) {
exit_code = WRITE_CGROUP_FAILED; exit_code = WRITE_CGROUP_FAILED;
goto cleanup; goto post_launch_cleanup;
} }
} }
} }
@ -1532,7 +1532,7 @@ int launch_docker_container_as_user(const char * user, const char *app_id,
exit_code = WRITE_PIDFILE_FAILED; exit_code = WRITE_PIDFILE_FAILED;
fprintf(ERRORFILE, "Could not write pid to %s", pid_file); fprintf(ERRORFILE, "Could not write pid to %s", pid_file);
fflush(ERRORFILE); fflush(ERRORFILE);
goto cleanup; goto post_launch_cleanup;
} }
snprintf(docker_wait_command, command_size, snprintf(docker_wait_command, command_size,
@ -1578,20 +1578,49 @@ int launch_docker_container_as_user(const char * user, const char *app_id,
} }
} }
post_launch_cleanup:
fprintf(LOGFILE, "Removing docker container post-exit...\n"); fprintf(LOGFILE, "Removing docker container post-exit...\n");
snprintf(docker_rm_command, command_size, snprintf(docker_rm_command, command_size,
"%s rm %s", docker_binary, container_id); "%s rm %s", docker_binary, container_id);
FILE* rm_docker = popen(docker_rm_command, "w"); int rc, i, sleep_time = 1, max_iterations = 5;
if (pclose (rm_docker) != 0) for (i = 0; i < max_iterations; i++) {
{ if (i > 0) {
fprintf (ERRORFILE, sleep(sleep_time);
"Could not remove container %s.\n", docker_rm_command); sleep_time *= 2;
fflush(ERRORFILE); }
exit_code = UNABLE_TO_EXECUTE_CONTAINER_SCRIPT; FILE* rm_docker = popen(docker_rm_command, "w");
goto cleanup; if (rm_docker == 0) {
fprintf(ERRORFILE,
"popen() failed: %s\n", strerror(errno));
fflush(ERRORFILE);
continue;
}
rc = pclose(rm_docker);
if (rc == -1) {
fprintf(ERRORFILE,
"pclose() failed: %s\n", strerror(errno));
fflush(ERRORFILE);
} else if (WIFEXITED(rc)) {
if (WEXITSTATUS(rc) == 0) {
break;
} else {
fprintf(ERRORFILE,
"docker rm command failed with exit status: %d\n", WEXITSTATUS(rc));
fflush(ERRORFILE);
}
}
} }
cleanup: if (i == max_iterations) {
// Tried 5 times and failed.
fprintf(ERRORFILE,
"Could not remove container after %d tries: %s\n", max_iterations, docker_rm_command);
fflush(ERRORFILE);
exit_code = UNABLE_TO_EXECUTE_CONTAINER_SCRIPT;
}
pre_launch_cleanup:
if (exit_code_file != NULL && write_exit_code_file_as_nm(exit_code_file, exit_code) < 0) { if (exit_code_file != NULL && write_exit_code_file_as_nm(exit_code_file, exit_code) < 0) {
fprintf (ERRORFILE, fprintf (ERRORFILE,