慢ssh登录-激活org.freedesktop.login1超时

Question

在我的服务器上，我注意到SSH登录确实延迟了。

使用ssh -vvv选项连接时，延迟发生在debug1: Entering interactive session.

连接的摘录：

debug1: Authentication succeeded (publickey).
Authenticated to IP_REDACTED ([IP_REDACTED]:22).
debug1: channel 0: new [client-session]
debug3: ssh_session2_open: channel_new: 0
debug2: channel 0: send open
debug1: Requesting no-more-sessions@openssh.com
debug1: Entering interactive session.
debug2: callback start
debug2: fd 3 setting TCP_NODELAY
debug3: packet_set_tos: set IP_TOS 0x10
debug2: client_session2_setup: id 0
debug2: channel 0: request pty-req confirm 1

使用概述这里的方法，我生成了strace输出，并注意到行14:09:53.676004 ppoll([{fd=5, events=POLLIN}], 1, {24, 999645000}, NULL, 8) = 1 ([{fd=5, revents=POLLIN}], left {0, 0}) <25.020764>需要25秒。

斯特拉斯输出的提取：

14:09:53.675567 clock_gettime(CLOCK_MONOTONIC, {4662549, 999741404}) = 0 <0.000024>
14:09:53.675651 recvmsg(5, {msg_name(0)=NULL, msg_iov(1)=[{"l\4\1\1\n\0\0\0\2\0\0\0\215\0\0\0\1\1o\0\25\0\0\0", 24}], msg_controll
en=0, msg_flags=MSG_CMSG_CLOEXEC}, MSG_DONTWAIT|MSG_NOSIGNAL|MSG_CMSG_CLOEXEC) = 24 <0.000024>
14:09:53.675744 recvmsg(5, {msg_name(0)=NULL, msg_iov(1)=[{"/org/freedesktop/DBus\0\0\0\2\1s\0\24\0\0\0"..., 146}], msg_controllen
=0, msg_flags=MSG_CMSG_CLOEXEC}, MSG_DONTWAIT|MSG_NOSIGNAL|MSG_CMSG_CLOEXEC) = 146 <0.000025>
14:09:53.675842 recvmsg(5, 0x7ffe0ff1dfa0, MSG_DONTWAIT|MSG_NOSIGNAL|MSG_CMSG_CLOEXEC) = -1 EAGAIN (Resource temporarily unavailab
le) <0.000023>
14:09:53.675925 clock_gettime(CLOCK_MONOTONIC, {4662550, 96075}) = 0 <0.000024>
14:09:53.676004 ppoll([{fd=5, events=POLLIN}], 1, {24, 999645000}, NULL, 8) = 1 ([{fd=5, revents=POLLIN}], left {0, 0}) <25.020764>
14:10:18.696865 recvmsg(5, {msg_name(0)=NULL, msg_iov(1)=[{"l\3\1\0013\0\0\0\3\0\0\0m\0\0\0\6\1s\0\5\0\0\0", 24}], msg_controllen=0,     msg_flags=MSG_CMSG_CLOEXEC}, MSG_DONTWAIT|MSG_NOSIGNAL|MSG_CMSG_CLOEXEC) = 24 <0.000017>
14:10:18.696944 recvmsg(5, {msg_name(0)=NULL, msg_iov(1)=[{":1.10\0\0\0\4\1s\0#\0\0\0org.freedesktop."..., 155}], msg_controllen=0, msg_flags=MSG_CMSG_CLOEXEC}, MSG_DONTWAIT|MSG_NOSIGNAL|MSG_CMSG_CLOEXEC) = 155 <0.000018>

我注意到在相关的时候在auth日志中有一个条目：

Jul 21 14:10:18 click sshd[8165]: pam_systemd(sshd:session): Failed to create session: Activation of org.freedesktop.login1 timed out

对此还不太清楚，它想要投票的是什么，为什么现在在这个特定的服务器上花了25秒。

journalctl -u systemd-logind命令显示

Jul 20 11:33:06 click systemd-logind[19415]: Failed to abandon session scope: Transport endpoint is not connected
Jul 21 05:04:54 myhost systemd[1]: Started Login Service.
Jul 21 12:15:30 myhost systemd[1]: Started Login Service.
Jul 21 12:17:04 myhost systemd[1]: Started Login Service.
Jul 21 12:49:55 myhost systemd[1]: Started Login Service.
Jul 21 13:57:05 myhost systemd[1]: Started Login Service.
Jul 21 13:58:49 myhost systemd[1]: Started Login Service.
Jul 21 14:01:55 myhost systemd[1]: Started Login Service.
Jul 21 14:08:32 myhost systemd[1]: Started Login Service.
Jul 21 14:09:53 myhost systemd[1]: Started Login Service.
Jul 21 14:19:08 myhost systemd[1]: Started Login Service.
Jul 21 14:21:26 myhost systemd[1]: Started Login Service.
Jul 21 14:22:37 myhost systemd[1]: Started Login Service.
Jul 21 14:25:20 myhost systemd[1]: Started Login Service.
Jul 21 14:30:27 myhost systemd[1]: Started Login Service.
Jul 21 15:02:56 myhost systemd[1]: Started Login Service.

发出命令systemctl restart systemd-logind.service修复它(目前可能是这样)。

它提到的Activation of org.freedesktop.login1是什么？有什么办法可以避免以后不得不重新启动logind吗？我希望随着时间的推移，我管理的其他服务器都会出现这个问题。

刚刚注意到这开始发生在另一台服务器上。

$ sudo service systemd-logind status

● systemd-logind.service - Login Service
   Loaded: loaded (/lib/systemd/system/systemd-logind.service; static)
   Active: active (running) since Tue 2015-06-16 14:10:57 BST; 1 months 12 days ago
     Docs: man:systemd-logind.service(8)
           man:logind.conf(5)
           http://www.freedesktop.org/wiki/Software/systemd/logind
           http://www.freedesktop.org/wiki/Software/systemd/multiseat
 Main PID: 1701 (systemd-logind)
   Status: "Processing requests..."
   CGroup: /system.slice/systemd-logind.service
           └─1701 /lib/systemd/systemd-logind

Jul 28 13:16:21 myhost systemd[1]: Started Login Service.
Jul 28 13:16:47 myhost systemd[1]: Started Login Service.
Jul 28 16:09:23 myhost systemd[1]: Started Login Service.
Jul 28 16:09:49 myhost systemd[1]: Started Login Service.
Jul 28 16:10:15 myhost systemd[1]: Started Login Service.
Jul 28 16:10:41 myhost systemd[1]: Started Login Service.
Jul 28 22:50:19 myhost systemd[1]: Started Login Service.
Jul 29 05:00:15 myhost systemd[1]: Started Login Service.
Jul 29 11:00:20 myhost systemd[1]: Started Login Service.
Jul 29 11:09:56 myhost systemd[1]: Started Login Service.

编辑-扩展的journalctl输出。

当注意到在另一台服务器上启动时，EDIT2添加了systemd-logind状态，如注释中所建议的。

更新-这开始发生在我的其他杰茜服务器。我是唯一一个经历这种事的人吗？除了重新启动系统之外，还有别的办法吗?有谁想过吗？

这个770135上有一个Debian报告。

Answer 1

当dbus重新启动，而systemd没有重新启动时，就会发生这种情况。只需做以下几点：

systemctl restart systemd-logind

解决方案来自于这里：https://major.io/2015/07/27/very-slow-ssh-logins-on-fedora-22/

Answer 2

使用：

systemctl restart systemd-logind

只能暂时解决问题。

解决方法之一是从cron作业中删除所有.scope文件，如声明的这里。

* 2,14 * * * root /bin/rm -f /run/systemd/system/*.scope

相关的systemd bug报告在这里：范围单元泄漏减慢"systemctl列表-单元-文件“和延迟登录。

它似乎实际上是一个dbus bug：unix在飞计数中断，它在dbus版本1.11.10中得到解决。

要永久修复此错误，只需等待dbus的此版本出现在您的发行版中即可。目前，Debian拉伸位于dbus 1.10.18，Ubuntu17.04(Z校)位于1.10.10，CentOS 7位于dbus 1.6.12。