主页 / unix / 问题 / 717642
Accepted
Fonic
Asked: 2022-09-18 06:18:07 +0800 CST

如何让 systemd 服务重新启动其他服务？

  • 772

我有一个应用程序，它在启动时从数据库中读取，该数据库由单独的 updater每天更新，由以下systemd单元描述：

application.service：

[Unit]
Description=Application reading database
Wants=network-online.target
After=local-fs.target network-online.target nss-lookup.target

[Service]
User=application
Group=application
WorkingDirectory=/var/lib/application
ExecStart=/var/lib/application/application-exec --database /var/db/database.db
StandardOutput=file:/var/lib/application/application.stdout.log
StandardError=file:/var/lib/application/application.stderr.log

StartLimitInterval=60
StartLimitBurst=10
Restart=always
RestartSec=10

[Install]
WantedBy=multi-user.target

updater.service：

[Unit]
Description=Database updater
Wants=network-online.target
After=local-fs.target network-online.target nss-lookup.target

[Service]
Type=oneshot
User=updater
Group=updater
WorkingDirectory=/var/lib/updater
ExecStart=/var/lib/updater/updater-exec --database /var/db/database.db
StandardOutput=file:/var/lib/updater/updater.stdout.log
StandardError=file:/var/lib/updater/updater.stderr.log

[Install]
WantedBy=multi-user.target

updater.timer：

[Unit]
Description=Timer for database updater

[Timer]
OnCalendar=*-*-* 03:00:00
RandomizedDelaySec=900
Persistent=true

[Install]
WantedBy=timers.target

我的问题：

如果更新成功执行，如何让更新服务重新启动应用服务（以便应用重新读取现在更新的数据库；成功/失败由更新的退出代码指示）？

这个关于serverfault的问题类似，但遗憾的是没有适用于我的用例的答案。

systemd services

2 个回答

  1. 这是一个有趣的发现！

    我们将向组合中添加第三个服务restart-application.service，然后安排和之间的依赖关系restart-applicationservice，updater.service这样：

    • 启动restart-application.service开始updater.service
    • restart-application.service直到updater.service完成才真正运行。
    • restart-application.serviceupdater.service除非成功，否则不会运行。

    我们restart-application.service这样创建：

    [Unit]
Description=Restart application
BindsTo=updater.service
After=updater.service

[Service]
Type=oneshot
ExecStart=/usr/bin/systemctl restart application

    我们在BindsTo这里使用依赖项。文档的关键部分是：

    当在同一单元上与 After= 结合使用时，BindsTo= 的行为会更加强大。在这种情况下，严格绑定到的单元必须处于活动状态才能使该单元也处于活动状态。

    现在我们将 的Unit部分修改updater.service为如下所示：

    [Unit]
Description=Database updater
Wants=network-online.target
After=local-fs.target network-online.target nss-lookup.target
StopWhenUnneeded=true

    以及要包括的Service部分：

    RemainAfterExit=true

    这些更改意味着 (a)在命令完成updater.service后将保持“活动” ；ExecStart我们需要这个，因为restart-application.service只有当这个处于活动状态时才会运行。该StopWhenUnneeded=true选项意味着当没有其他东西依赖它时（即restart-application.service停止时），该单元将变为非活动状态。

    最后，我们需要修改您的计时器以启动restart-application.service而不是updater.service.

    我创建了一个假人application.service（只运行一个简单的网络服务器）和一个updater.service随机成功或失败的假人：

    [Unit]
Description=Database updater
Wants=network-online.target
After=local-fs.target network-online.target nss-lookup.target
StopWhenUnneeded=true

[Service]
Type=oneshot
User=updater
Group=updater
WorkingDirectory=/home/updater
ExecStart=/bin/bash -c 'rc=$(( RANDOM % 2 )); echo "rc=$rc"; exit $rc'
RemainAfterExit=true

    让我们在systemctl start restart-application.service. 这是updater.service失败时会发生的情况；

    ... systemd[1]: Starting updater.service - Database updater...
... bash[1253]: rc=1
... systemd[1]: updater.service: Main process exited, code=exited, status=1/FAILURE
... systemd[1]: updater.service: Failed with result 'exit-code'.
... systemd[1]: Failed to start updater.service - Database updater.
... systemd[1]: Dependency failed for restart-application.service - Restart application.
... systemd[1]: restart-application.service: Job restart-application.service/start failed with result 'dependency'.

    updater.service以下是成功时会发生的情况：

    ... systemd[1]: Starting updater.service - Database updater...
... systemd[1]: Finished updater.service - Database updater.
... darkhttpd[1193]: darkhttpd/1.13, copyright (c) 2003-2021 Emil Mikulic.
... darkhttpd[1193]: listening on: http://0.0.0.0:8080/
... systemd[1]: Starting restart-application.service - Restart application...
... bash[1262]: rc=0
... systemd[1]: Stopping application.service - Application reading database...
... systemd[1]: application.service: Deactivated successfully.
... systemd[1]: Stopped application.service - Application reading database.
... systemd[1]: Started application.service - Application reading database.
... systemd[1]: restart-application.service: Deactivated successfully.
... systemd[1]: Finished restart-application.service - Restart application.
... systemd[1]: updater.service: Deactivated successfully.
... systemd[1]: Stopped updater.service - Database updater.

    此时一切的状态：

    # systemctl status application.service updater.service restart-application.service
● application.service - Application reading database
     Loaded: loaded (/etc/systemd/system/application.service; disabled; vendor preset: disabled)
     Active: active (running) since Sat 2022-09-17 20:48:27 UTC; 1min 53s ago
   Main PID: 1264 (darkhttpd)
      Tasks: 1 (limit: 4661)
     Memory: 172.0K
        CPU: 1ms
     CGroup: /system.slice/application.service
             └─ 1264 /usr/sbin/darkhttpd /home/application/htdocs --port 8080

○ updater.service - Database updater
     Loaded: loaded (/etc/systemd/system/updater.service; disabled; vendor preset: disabled)
     Active: inactive (dead)

○ restart-application.service - Restart application
     Loaded: loaded (/etc/systemd/system/restart-application.service; static)
     Active: inactive (dead)

    我认为这可以满足您的所有需求。

    • 1
  2. Best Answer

    结果证明解决方案比预期的要简单得多。以下行需要添加到updater.service：

    ExecStartPost=+/usr/bin/systemctl restart application.service

    结果updater.service：

    [Unit]
Description=Database updater
Wants=network-online.target
After=local-fs.target network-online.target nss-lookup.target

[Service]
Type=oneshot
User=updater
Group=updater
WorkingDirectory=/var/lib/updater
ExecStart=/var/lib/updater/updater-exec --database /var/db/database.db
StandardOutput=file:/var/lib/updater/updater.stdout.log
StandardError=file:/var/lib/updater/updater.stderr.log
ExecStartPost=+/usr/bin/systemctl restart application.service

[Install]
WantedBy=multi-user.target

    相关摘录man 5 systemd.service：

    ├───────┼────────────────────────────────────────┤
│"+"    │ If the executable path is prefixed     │
│       │ with "+" then the process is executed  │
│       │ with full privileges. In this mode     │
│       │ privilege restrictions configured with │
│       │ User=, Group=, CapabilityBoundingSet=  │
│       │ or the various file system namespacing │
│       │ options (such as PrivateDevices=,      │
│       │ PrivateTmp=) are not applied to the    │
│       │ invoked command line (but still affect │
│       │ any other ExecStart=, ExecStop=, ...   │
│       │ lines).                                │
├───────┼────────────────────────────────────────┤

    ExecStartPost= commands are only run after the commands specified in ExecStart= have been invoked
successfully, as determined by Type= (i.e. the process has been started for Type=simple or Type=idle,
the last ExecStart= process exited successfully for Type=oneshot, the initial process exited
successfully for Type=forking, "READY=1" is sent for Type=notify, or the BusName= has been taken for
Type=dbus).

    解释：

    • ExecStartPost=由于前面的原因，命令以 root 权限执行（+参见第一个摘录），尽管指定和application.servicesystemctlUser=Group=
    • ExecStartPost=只有在命令成功时才会执行ExecStart=命令（参见第二个摘录），这允许有条件地重新启动application.service
    • 1

相关问题