Working with VMware Bitfusion Application Logs

Products

• VMware
• VMware vSphere ESXi

Issue/Introduction

VMware Bitfusion produces logs files, traces, and statistics for troubleshooting. The log level can be set by users and administrators. Log files are created by default; you must enable the creation of any additional files. The “capture” command option is the most convenient way to do this.

Environment

VMware vSphere Bitfusion 3.x
VMware vSphere Bitfusion 4.x
VMware vSphere Bitfusion 2.x

Resolution

This solution is separated into two sections:

• Logging
• Understanding the capture command

Logging

The available log levels are:

• TRACE: the most-verbose logging level with maximum information
• INFO: Includes errors and informational messages
• WARN: Logs issues, but not other information
• ERROR: Logs error messages only
• FATAL: Logs only conditions that result in application failure

 

To adjust the log level permanently use the export command with either or both LEVEL variables:
export <variable>=<level>

Variable	Available Options	Purpose
BF_LOG_LEVEL	TRACE|INFO|WARN|ERROR|FATAL	The default global log level
BF_LOG_LEVEL_rpc_api	TRACE|INFO|WARN|ERROR|FATAL	RPC API subsystem log level

For example, the command to set the global log level to "INFO" is:
export BF_LOG_LEVEL=INFO
 

Logs are kept in the .bitfusion subdirectory of the user’s home and in the system log:
~/.bitfusion/bf_Global.log
and
/var/log/syslog (may vary depending on OS or version)

The variables will be read and used when Bitfusion is invoked and not mid-run, so the Bitfusion service will have to be restarted after adjusting the logging variables on any Bitfusion GPU server. This will cause a temporary interruption in Bitfusion services:
# systemctl restart bitfusion

 

Understanding the capture command

Bitfusion has a --capture option to the run command that can gather logs, traces, statistics and core dumps into a tarfile; and it lets you set the log level without explicitly setting the environmental variables discussed above.  Here is an example:
bitfusion run –n 1 –-capture stats,tracing,logs:WARN ...

 

The capture option works as follows:
--capture <value>:<log level>

• value can be a comma-separated list (no whitespace) of the following key words
  • logs, tracing, stats, and coredump.
  • Alternately all can be selected by using the keyword: all.
• log level can take a specifier to set the log level:
  • ERROR, WARN, INFO, TRACE, DEBUG.
• Note: If you use all for the value, you cannot set the log level.