ClearML/clearml-agent
1
0
Fork 0
mirror of https://github.com/clearml/clearml-agent synced 2025-03-03 18:52:22 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Change daemon --stop without any specific flag to terminate the agents by worker id lexicographic order

Browse Source
This commit is contained in:
allegroai 2020-07-11 01:43:54 +03:00
parent 94997f9c88
commit 17d2bf2a3e
2 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
trains_agent/commands/worker.py
View File

@ -707,9 +707,10 @@ class Worker(ServiceCommandSection):
kwargs = self._verify_command_states(kwargs) kwargs = self._verify_command_states(kwargs)
docker = docker or kwargs.get('docker') docker = docker or kwargs.get('docker')
# We are not running a daemon we are killing one.
# find the pid send termination signal and leave
if kwargs.get('stop', False): if kwargs.get('stop', False):
self._kill_daemon() return 1 if not self._kill_daemon() else 0
return
# make sure we only have a single instance, # make sure we only have a single instance,
# also make sure we set worker_id properly and cache folders # also make sure we set worker_id properly and cache folders
@ -2355,7 +2356,7 @@ class Worker(ServiceCommandSection):
def _kill_daemon(self): def _kill_daemon(self):
worker_id, worker_name = self._generate_worker_id_name() worker_id, worker_name = self._generate_worker_id_name()
# Iterate over all running process # Iterate over all running process
for pid, uid, slot, file in Singleton.get_running_pids(): for pid, uid, slot, file in sorted(Singleton.get_running_pids(), key=lambda x: x[1] or ''):
# wither we have a match for the worker_id or we just pick the first one # wither we have a match for the worker_id or we just pick the first one
if pid >= 0 and ( if pid >= 0 and (
(worker_id and uid == worker_id) or (worker_id and uid == worker_id) or

1
trains_agent/helper/process.py
View File

@ -65,6 +65,7 @@ def terminate_process(pid, timeout=10.):
except Exception: except Exception:
return True return True
def kill_all_child_processes(pid=None): def kill_all_child_processes(pid=None):
# get current process if pid not provided # get current process if pid not provided
include_parent = True include_parent = True