有没有办法知道为什么重新启动服务以及是谁做的？

Question

• Ubuntu 14.04
• 克拉玛夫 0.98.7

问题是clamav-daemon几乎每天都重新启动：

Sep  1 06:30:00 x-master clamd[6778]: Pid file removed.
clamd[6778]: --- Stopped at Tue Sep  1 06:30:00 2015
clamd[5979]: clamd daemon 0.98.7 (OS: linux-gnu, ARCH: x86_64, CPU: x86_64)
clamd[5979]: Running as user root (UID 0, GID 0)
clamd[5979]: Log file size limited to 4294967295 bytes.
clamd[5979]: Reading databases from /var/lib/clamav
clamd[5979]: Not loading PUA signatures.
clamd[5979]: Bytecode: Security mode set to "TrustSigned".

如果clamdscan正在运行，则会导致问题：

/etc/cron.daily/clamav_scan:
ERROR: Could not connect to clamd on x.x.x.x: Connection refused

请注意，我在开头说的是“几乎”：

/var/log/syslog:Sep  1 06:30:00 x-master clamd[6778]: Pid file removed.
/var/log/syslog.1:Aug 31 06:27:54 x-master clamd[20128]: Pid file removed.
/var/log/syslog.4.gz:Aug 28 06:28:34 x-master clamd[4475]: Pid file removed.
/var/log/syslog.5.gz:Aug 27 06:27:47 x-master clamd[21466]: Pid file removed.

如你看到的：

• 它没有发生在 8 月 29 日和 30 日

• 它经常在 06:27 左右重新启动，这cron.daily是运行时间

  27 6 * * * root nice -n 19 ionice -c3 run-parts --report /etc/cron.daily
  

内容/etc/cron.daily/clamav_scan：

find / $exclude_string ! \( -path "/tmp/clamav-*.tmp" -prune \) ! \( -path "/var/lib/elasticsearch" -prune \) ! \( -path "/var/lib/mongodb" -prune \) ! \( -path "/var/lib/graylog-server" -prune \) -mtime -1 -type f -print0 | xargs -0 clamdscan --quiet -l "$status_file" || retval=$?

clamav-daemon 有一个 logrotate 文件：

/var/log/clamav/clamav.log {
     rotate 12
     weekly
     compress
     delaycompress
     create 640  clamav adm
     postrotate
     /etc/init.d/clamav-daemon reload-log > /dev/null
     endscript
     }

但它只是重新加载日志：

Sep  1 02:30:24 uba-master clamd[6778]: SIGHUP caught: re-opening log file.

我知道我们可以auditd用来监控二进制文件，这里有一个示例日志：

ausearch -f /usr/sbin/clamd                                                                                                        [2/178]
----
time->Tue Sep  1 07:56:44 2015
type=PATH msg=audit(1441094204.559:15): item=1 name=(null) inode=2756458 dev=fc:00 mode=0100755 ouid=0 ogid=0 rdev=00:00
type=PATH msg=audit(1441094204.559:15): item=0 name="/usr/sbin/clamd" inode=3428628 dev=fc:00 mode=0100755 ouid=0 ogid=0 rdev=00:00
type=CWD msg=audit(1441094204.559:15):  cwd="/"
type=EXECVE msg=audit(1441094204.559:15): argc=1 a0="/usr/sbin/clamd"
type=SYSCALL msg=audit(1441094204.559:15): arch=c000003e syscall=59 success=yes exit=0 a0=7ffd277e03dc a1=7ffd277dfa78 a2=7ffd277dfa88 a3=7ffd277df570 items=2
 ppid=5708 pid=5946 auid=4294967295 uid=109 gid=114 euid=109 suid=109 fsuid=109 egid=114 sgid=114 fsgid=114 tty=pts1 ses=4294967295 comm="clamd" exe="/usr/sbin/clamd" key=(null)

109是...clamav用户的UID ：

getent passwd clamav clamav:x:109:114::/var/lib/clamav:/bin/false

在这种情况下，还有其他方法可以排除故障吗？

---

回复@HBruijn：

更新AV定义后可能是freshclam？

我想过这个问题。这是日志：

Sep  1 05:31:04 x-master freshclam[16197]: Received signal: wake up
Sep  1 05:31:04 x-master freshclam[16197]: ClamAV update process started at Tue Sep  1 05:31:04 2015
Sep  1 05:31:04 x-master freshclam[16197]: main.cvd is up to date (version: 55, sigs: 2424225, f-level: 60, builder: neo)
Sep  1 05:31:05 x-master freshclam[16197]: Downloading daily-20865.cdiff [100%]
Sep  1 05:31:09 x-master freshclam[16197]: daily.cld updated (version: 20865, sigs: 1555338, f-level: 63, builder: neo)
Sep  1 05:31:10 x-master freshclam[16197]: bytecode.cvd is up to date (version: 268, sigs: 47, f-level: 63, builder: anvilleg)
Sep  1 05:31:13 x-master freshclam[16197]: Database updated (3979610 signatures) from db.local.clamav.net (IP: 168.143.19.95)
Sep  1 05:31:13 x-master freshclam[16197]: Clamd successfully notified about the update.
Sep  1 05:31:13 x-master freshclam[16197]: --------------------------------------
Sep  1 04:34:10 x-master clamd[6778]: SelfCheck: Database status OK.
Sep  1 05:31:13 x-master clamd[6778]: Reading databases from /var/lib/clamav
Sep  1 05:31:22 x-master clamd[6778]: Database correctly reloaded (3974071 signatures)

我不确定这一点，但看起来 freshclam 有一个“内部机制”来通知 clamd 更新。之后它可以重新加载数据库，无需重新启动该过程。你可否确认？

此外，从时间戳中，我看到在freshclam 更新数据库一小时后，clamav-daemon 重新启动。正常吗？

---

更新 2015 年 9 月 1 日星期二 22:10:49 ICT

但看起来freshclam 有一个“内部机制”来通知clamd 更新。之后它可以重新加载数据库，无需重新启动该过程。

我可以通过进行测试来确认这是正确的：

• 编辑 freshclam.conf 文件以将间隔更改为分钟 ( Checks 1440)
• 重启 clamav-freshclam
• cd /var/lib/clamav
• rm每天.cvd

• 等一分钟

  Sep  1 14:49:25 p freshclam[7654]: Downloading daily.cvd [100%]
  Sep  1 14:49:28 p freshclam[7654]: daily.cvd updated (version: 19487, sigs: 1191913, f-level: 63, builder: neo)
  Sep  1 14:49:28 p freshclam[7654]: Reading CVD header (bytecode.cvd):
  Sep  1 14:49:28 p freshclam[7654]: OK
  Sep  1 14:49:28 p freshclam[7654]: bytecode.cvd is up to date (version: 245, sigs: 43, f-level: 63, builder: dgoddard)
  Sep  1 14:49:31 p freshclam[7654]: Database updated (3616181 signatures) from clamav.local (IP: 10.0.2.2)
  Sep  1 14:49:31 p freshclam[7654]: Clamd successfully notified about the update.
  Sep  1 14:49:31 p freshclam[7654]: --------------------------------------
  Sep  1 14:49:32 p clamd[6693]: Reading databases from /var/lib/clamav
  Sep  1 14:49:39 p clamd[6693]: Database correctly reloaded (3610621 signatures)
  

并且 clamav-daemon 没有重新启动。

Answer 1

请检查您是否使用了任何配置管理系统，例如 Puppet、Chef、CFEngine 等。它们会定期干扰服务。为纠正此问题而采取的确切措施取决于该服务在配置管理系统中的使用方式。

Answer 2

注意我自己。

作业缓存的输出：

----------
          ID: clamav-daemon
    Function: service.running
      Result: True
     Comment: Service restarted
     Started: 06:27:52.736890
    Duration: 12997.632 ms
     Changes:
              ----------
              clamav-daemon:
                  True

看一下clamav公式：

  clamav-daemon:
    service:
      - running
      - order: 50
      - require:
        - service: clamav-freshclam
      - watch:
        - pkg: clamav-daemon
        - file: clamav-daemon
        - user: clamav

watched 状态中没有任何更改：

----------
          ID: clamav-daemon
    Function: pkg.latest
      Result: True
     Comment: Package clamav-daemon is already up-to-date.
     Started: 06:27:51.531415
    Duration: 53.224 ms
     Changes:

----------
          ID: clamav-daemon
    Function: file.managed
        Name: /etc/clamav/clamd.conf
      Result: True
     Comment: File /etc/clamav/clamd.conf is in the correct state
     Started: 06:27:51.760019
    Duration: 625.075 ms
     Changes:

----------
          ID: clamav
    Function: user.present
      Result: True
     Comment: User clamav is present and up to date
     Started: 06:27:51.590214
    Duration: 2.455 ms
     Changes:

为什么服务已经重启？

寻找watch_in，我发现了一个管理pid文件的状态，如果pid文件发生变化，服务将重新启动：

{%- macro manage_pid(path, user, group, watch_in_service, mode=644) -%}
    {%- if salt['file.file_exists'](path) %}
{{ path }}:
  file:
    - managed
    - user: {{ user }}
    - group: {{ group }}
    - mode: {{ mode }}
    - replace: False
        {%- if caller is defined -%}
            {%- for line in caller().split("\n") -%}
                {%- if loop.first %}
    - require:
                {%- endif %}
{{ line|trim|indent(6, indentfirst=True) }}
            {%- endfor -%}
        {%- endif %}
    - watch_in:
      - service: {{ watch_in_service }}
    {%- else %}
# {{ path }} does not exist, no need to manage
    {%- endif -%}
{%- endmacro -%}

{%- call manage_pid('/var/run/clamav/clamd.pid', 'clamav', 'clamav', 'clamav-daemon', 664) %}
- pkg: clamav-daemon
{%- endcall %}

在 的输出中salt-run jobs.lookup_jid <job id number>，我看到了：

----------
          ID: /var/run/clamav/clamd.pid
    Function: file.managed
      Result: True
     Comment:
     Started: 06:27:52.392555
    Duration: 2.364 ms
     Changes:
              ----------
              group:
                  clamav
              user:
                  clamav

因此，该 pid 文件的所有者/组已更改为clamav. 最后，我发现原因是 clamav 守护进程以root用户身份在网络模式下运行。因此，pid 文件被创建为root. 因此，管理 pid 文件的状态必须更改为：

{%- call manage_pid('/var/run/clamav/clamd.pid', 'root', 'root', 'clamav-daemon', 664) %}
- pkg: clamav-daemon
{%- endcall %}