Logging in the library and the daemon

Libvirt includes logging facilities starting from version 0.6.0, this complements the error handling mechanism and APIs to allow tracing through the execution of the library as well as in the libvirtd daemon.

Logging in the library

The logging functionalities in libvirt are based on 3 key concepts, similar to the one present in other generic logging facilities like log4j:

Note that the logging module saves all logs to a debug buffer filled in a round-robin fashion as to keep a full log of the recent logs including all debug. The debug buffer can be resized or deactivated in the daemon using the log_buffer_size variable, default is 64 kB. This can be used when debugging the library (see the virLogBuffer variable content).

Configuring logging in the library

The library configuration of logging is through 3 environment variables allowing to control the logging behaviour:

Note that, for example, setting LIBVIRT_DEBUG= is the same as unset. If you specify an invalid value, it will be ignored with a warning. If you have an error in a filter or output string, some of the settings may be applied up to the point at which libvirt encountered the error.

Logging in the daemon

Similarly the daemon logging behaviour can be tuned using 3 config variables, stored in the configuration file:

When starting the libvirt daemon, any logging environment variable settings will override settings in the config file. Command line options take precedence over all. If no outputs are defined for libvirtd, it defaults to logging to /var/log/libvirt/libvirtd.log (before 0.9.0 it was using syslog) when it is running as a daemon, or to stderr when it is running in the foreground.

Libvirtd does not reload its logging configuration when issued a SIGHUP. If you want to reload the configuration, you must do a service libvirtd restart or manually stop and restart the daemon yourself.

Starting from 0.9.0, the daemon can save all the content of the debug buffer to the defined error channels (or /var/log/libvirt/libvirtd.log by default) in case of crash, this can also be activated explicitly for debugging purposes by sending the daemon a USR2 signal:

killall -USR2 libvirtd

Syntax for filters and output values

The syntax for filters and outputs is the same for both types of variables.

The format for a filter is one of:

  x:name  (log message only)
  x:+name (log message + stack trace)

where name is a string which is matched against source file name, e.g., remote, qemu, or util/json, the optional + prefix tells libvirt to log stack trace for each message matching name, and x is the minimal level where matching messages should be logged:

Multiple filters can be defined in a single string, they just need to be separated by spaces, e.g: "3:remote 4:event" to only get warning or errors from the remote layer and only errors from the event layer.

If you specify a log priority in a filter that is below the default log priority level, messages that match that filter will still be logged, while others will not. In order to see those messages, you must also have an output defined that includes the priority level of your filter.

The format for an output can be one of those 3 forms:

In all cases the x prefix is the minimal level, acting as a filter:

Multiple output can be defined, they just need to be separated by spaces, e.g.: "3:syslog:libvirtd 1:file:/tmp/libvirt.log" will log all warnings and errors to syslog under the libvirtd ident but also log all debug and information included in the file /tmp/libvirt.log

Examples

For example setting up the following:

export LIBVIRT_DEBUG=1
export LIBVIRT_LOG_OUTPUTS="1:file:virsh.log"

and then running virsh will accumulate the logs in the virsh.log file in a way similar to:

14:29:04.771: debug : virInitialize:278 : register drivers
14:29:04.771: debug : virRegisterDriver:618 : registering Test as driver 0

the messages are timestamped, there is also the level recorded, if debug the name of the function is also printed and then the formatted message. This should be sufficient to at least get a precise idea of what is happening and where things are going wrong, allowing to then put the correct breakpoints when running under a debugger.

To activate full debug of the libvirt entry points, utility functions and the QEmu/KVM driver, set:

log_filters="1:libvirt 1:util 1:qemu"
log_outputs="1:file:/var/log/libvirt/libvirtd.log"

in libvirtd.conf and restart the daemon will allow to gather a copious amount of debugging traces for the operations done in those areas.

On the other hand to deactivate the logbuffer in the daemon for stable high load servers, set

log_buffer_size=0

in the libvirtd.conf.