Differences

This shows you the differences between two versions of the page.

--- public:trouble:libvirtd [2017/01/12 10:49] – [Failure cause] wiki.manager
+++ public:trouble:libvirtd [2017/01/12 10:54] (current) – [Failure cause] wiki.manager
@@ Line 1: / Line 1: @@
+===== Hanging process 'libvirtd' =====
+==== Error description ====
+A running daemon 'libvirtd' is the process through which the command 'virsh' communicates with running vhosts.
+Sometimes the process becomes inaccessible:
+  root@kvm55:~# ps ax |grep libvirtd
+?        Zl     0:00 [libvirtd] <defunct>
+As 'libvirtd' is inaccessible, no virsh command will be properly executed: it will hang forever. Thus all 'virsh' commands have to be avoided. Otherwise lots of hanging threads are generated. As a process libvirtd is running, no additional process 'libvirtd' can be started.
+==== Failure cause ====
+The error cause is unknown. It was once observed after an auto-restart of the service 'libvirt-bin' due to an auto-update of the package 'libvirt-bin'.
+Administration experience on Systems with Ubuntu 14.04 indicated, that a hanging libvirtd is more likely when libvirt-bin is restarted like so:
+  service libvirt-bin restart
+There are no experiences with Ubuntu 16.04 yet regarding this error.
+==== Solution ====
+To avoid the error on manual restart of libvirt-bin, at least on Ubuntu 14.04 restart libvirt-bin like so:
+  /etc/init.d/libvirt-bin restart
+Hanging processes can be killed by killing their parent process unless the parent process is the init (1) process:
+  root@kvm55:~# ps -ef |grep [d]efunct
+  UID        PID  PPID  C STIME TTY          TIME CMD
+  root      7691     1  0 Jul21 ?        00:00:00 [libvirtd] <defunct>
+The only known solution is a reboot of the KVM host thus interrupting all running vhosts.
+The vhosts remain running while 'libvirtd' is inaccessible. Thus the reboot can be postponed to a time when the vhosts are not urgently needed. In the meantime avoid every action creating massive numbers of hanging virsh commands like puppet agents and backup processes.
+During reboot of the kvm host the downtime for a vhost can be shortened by the following steps:
+  - Define the vhost on another KVM host with access to the virtual disk(s) of the vhost.
+  - Shut down the vhost from inside the vhost.
+  - Once the vhost is down start the vhost on the other KVM host.
+[[public:trouble:vmbuilder|previous chapter]] | [[public:start|contents]] | [[public:trouble:blockdev|next chapter]]