APC AP9212 User's Guide Page 84

  • Download
  • Add to my manuals
  • Print
  • Page
    / 147
  • Table of contents
  • BOOKMARKS
  • Rated. / 5. Based on customer reviews
Page view 83
84
10.3 Failure and fault analysis
As the complexity of any system grows, so does the likelihood of a individual components failure.
The consequences of this failure are unique to every system. This highlights the need to analyse
the system and identify faults that can occur and how these faults will propagate through the
system, on both a hardware and a software level. The following questions should be considered
when designing a control system:
What potential faults can occur and how serious are they?
How can software accurately become aware of these faults ?
What action can be taken to recover ?
How will this fault propagate to other devices in the System ?
Can additional hardware be added to the system to more effectively diagnose these
faults ?
This information would encourage the development of a more reliable software system capable
of detecting and recovering from failures.
10.4 Logging analysis
10.4.1 Overview
In virtually every complex software system, various log files are used for monitoring and
debugging the software. This characteristic is shared by both the APT and the Spectrum Scanner
software through the use of the java.util.logging framework. Operating Systems are also an
example of this. Unix/Linux based operating systems log system messages to
/var/adm/messages
i
whereas WindowsNT messages can viewed with the Event Viewer.
In any software failure, analysis of these log files is generally the standard procedure for
identifying the cause of the problem. Particularly when the system is controlled remotely. This
analysis can often be very time consuming due to the verbosity and sheer size of log information.
Command line tools such as grep, tail, more, and less are often used in this kind of
analysis. These are inadequate, particularly for Java based systems, for several reasons:
Log messages can be on more than one line (eg Java stack traces).
Some log messages are longer than 80 characters.
Log messages cannot be filtered according to their severity.
Searches are limited to plain text as opposed to log record types and attributes.
Given these shortcomings of the current log analysis tools, a specification for an ideal tool can
be defined.
10.4.2 Analysis tool specification
The aim is to provide a graphical based tool to analyse log file data using searching, filtering and
masking techniques. The power of this tool is drawn from the fact that LogRecord data types
can be searched, rather than just plain text.
i
The path of the messages log file varies between differnts distrubutions
Page view 83
1 2 ... 79 80 81 82 83 84 85 86 87 88 89 ... 146 147

Comments to this Manuals

No comments