systemd.timerがactivating状態のまま

現象

active

inactive

activating

deactivating

failed

Units may be "active" (meaning started, bound, plugged in, …, depending on the unit type, see below), or "inactive" (meaning stopped, unbound, unplugged, …), as well as in the process of being activated or deactivated, i.e. between the two states (these states are called "activating", "deactivating").

バッチの起動中になにかが詰まった可能性がある

例えば、バッチでメールサーバーにアクセスしているなら、メールサーバーからの応答が返ってこない、非常に遅い、など

タイムアウト設定できるの？

>> 2) What is systemd's timeout by default for service activation

>> (timer-activated, socket-activated)? If it is documentened, please, give

>> me a hint.

>

> By default, Type=oneshot services don't have a start timeout.

>> 3) If systemd's timeout from 2) is present, how can it be managed by

>> user\admin? E.g. after 10 minutes of 'Activating (start)', service gets

>> FAILED state with putting this info to systemd log, of course (something

>> ExitCode:<number>").

>

> If you do add a TimeoutStartSec= to a Type=oneshot unit, this will

> force the unit to be stopped (by killing the process) after that time.

> That's probably not what you want.

In general way -- why not? For more twisted cases I can use OnFailure=

directive here, I suppose. So, will try to play with TimeoutStartSec=

directive.

> For a Type=simple service, the default TimeoutStartSec= is set in

> /etc/systemd/system.conf. It will be 90 seconds unless you've changed it.

タイムアウトを設定できるらしい

ここまでのまとめ

その先の話

バッチのステータスを細かく知りたい場合は？

ログを出すといいのでは

バッチが長時間かかって、4分で終わらなくなったらどうするの？

Timeout方式ではなく、Watchdog方式で応答なくなったらKILLなどできないかな？

バッチ自身で、4分くらい実行したら終了するようにして、次の起動で続きの動作をする等はできないか？

WatchDog

時間を秒数で設定。その時間のあいだ通知がこなければ、発動される

Notify

サンプルコード

code:batch.service

# /etc/systemd/system/batch.service

User=ubuntu

Group=ubuntu

ExecStart=/usr/bin/python3 -u /home/ubuntu/systemdtest/batch.py

ExecStop=/bin/kill -s TERM $MAINPID

Type=notify # systemdのstatusをbatch.pyから制御

WatchdogSec=7 # batch.pyから7秒間 WATCHDOG=1 が来ない場合KILLする

Description=batch.py

After=multi-user.target

code:batch.timer

# /etc/systemd/system/batch.timer

Description=batch

OnBootSec=1min

OnUnitActiveSec=1min

WantedBy=timers.target

code:batch.py

import sys

import os

import socket

import time

import random

import signal

def notify(msg):

socket_path = os.environ.get('NOTIFY_SOCKET')

if not socket_path:

raise RuntimeError('No notify socket')

with socket.socket(socket.AF_UNIX, socket.SOCK_DGRAM) as s:

s.connect(socket_path)

s.sendall(msg.encode())

def loop():

for i in range(1, 11):

time.sleep(sec) # DO SOMETHING

print('... done')

notify('WATCHDOG=1') # update watchdog timestamp

notify('STATUS={} tasks are finished'.format(i)) # update systemd status

def initialize():

notify('STATUS=Process started. Setting up...') # update systemd status

signal.signal(signal.SIGABRT, handle_watchdog) # handle watchdog time limit

time.sleep(3)

notify('READY=1') # activating -> active

notify('STATUS=Ready.') # update systemd status

def handle_watchdog(signum, tb):

notify('STATUS=KILLED by watchdog') # update systemd status

# do some finialize process

sys.exit(-1) # inactive

def finalize():

notify('STATUS=Process closing...') # update systemd status

notify('STOPPING=1') # active -> deactivating

time.sleep(3) # do some finalize process

notify('STATUS=Finished.') # update systemd status

sys.exit(0) # inactive

def main():

initialize()

loop()

finalize()

if __name__ == '__main__':

main()

code:killed-in-activating.log

code:killed-in-active.log

ここまでのまとめ