Collecting diagnostic data for NFS

This topic describes the procedure for collecting diagnostic data for NFS services.

Diagnostic data can be generated by increasing the logging of the NFS server.

To change the logging temporarily on a single CES node without restarting the server, run the following the command on the CES node where you want to enable the tracing:

ganesha_mgr set log COMPONENT_ALL FULL_DEBUG

CES NFS log levels can be user adjusted to select the amount of logging by the server. Every increase in log setting will add additional messages. So the default of "Event" will include messages tagged as EVENT, WARN, CRIT, MAJ, FATAL, but will not show INFO, DEBUG, MID_DEBUG, FULL_DEBUG:

Table 1. CES NFS log levels
Log name	Description
NULL	No logging
FATAL	Only asserts are logged
MAJ	Only major events are logged
CRIT	Only critical events are logged where there is malfunction, that is, for a single request
WARN	Events are logged that may be intended but may otherwise be harmful
EVENT	Default. level, which includes some events that are expected during normal operation (that is, start, grace period),
INFO	Enhanced level
DEBUG	Further enhanced, which includes events-relevant problem determination
MID_DEBUG	Further enhanced, which includes some events for developers
FULL_DEBUG	Maximal logging, which is mainly used for development purposes

These levels can be applied to a single component or to all components.

Note: The ganesha_mgr command requires that the CES NFS server be active on the node where the command is executed.

The CES NFS log file (default is /var/log/ganesha.log) will see a lot more updates eventually generating very large files or even filling up all of your disk.

To avoid issues with space usage, revert to the default logging by using the ganesha_mgr set log COMPONENT_ALL EVENT command or reduce the set of components by using "FULL_DEBUG" to a reasonable subset of server components. For example, by replacing "COMPONENT_ALL" with "COMPONENT_DISPATCH".

Other possible components can be listed by using ganesha_mgr getall logs. The ganesha_mgr changes are not persistent. A server restart will reset these settings to the settings in the cluster configuration as described in the mmnfs config list command.

Note: The mmnfs config list will show the persisted log level for all CES nodes (default: EVENT). Any log setting changes by using mmnfs config change LOG_LEVEL command will automatically do a server restart, possibly preventing to find a cause for the current issue. See the mmnfs command command page for more information.