NetMon

Provide support for Continuous vs Discrete events. "process_event"
must check configuration file for respective attribute and decide
the action. For "Continuous" event types, no changes are proposed.
For "Discrete" event types, instead of doing "averaging" for 5
minues (or 30 minutes or 1 hour or 24 hours), keep 1st instance of
each possible values for a given time duration.

Discussion on 14 Nov 2006 (Shuvam, Kishan)

1. No averaging of the data for weekly, monthly and yearly graph.
(This is applicable for selected cases only).

Keep minimum and maximum value during the averaging period
and store both against the same time line (X axis).

Last priority among today's tasks.

2. Possible values for YMax

YMax = auto
= <val>
= PWR:<int> # Nearest power of this number > YMax
= MULT:<int> # Nearest multiple of this number > YMax

3. Ticks

$YLines = 'STEPS:<val>'

Indicates how many ticks to be drawn.

Add one more variale --

YLineDesc = " " # A template string which can be used within
# sprintf. First argument to sprintf will be
# the value

YLineDivisor = 1 # This can be used for nice, rounded descri

4. Agentd: Cycling of counts need to be fixed in the agent. In this
case, whenever a new value is < previous value, skip this value
for reporting to Netmon.

5. All traffic measurement agents report in units of kbits/sec
instead of Bytes/sec.

6. Netmon graph page

- For weekly graph, put "30-min average" e.g. space just like
other description.

- Move "report was last updated ..." to bottom of the page.

- Remove bold font in subsection description like
"Daily average", etc.

-x-

9 March 2007

Files changed so far:

1. /opt/merce/lib/perl/Linegraph.pm -- Draw the exception marks
above the total height of the graph.

2. /opt/merce/lib/perl/Emsg.pm -- Logs the IP address of the
incoming packet along with information sent by Netmon agent.

3. /opt/merce/bin/usr/agentd -- Added new agents for Windows server
running SNMP service, as well as "memfree" for Linux.

4. /opt/merce/doc/graphs/ -- Permissions and ownership. These are
important for proper functioning of Netmon.

5. /opt/merce/lib/perl/Rulebase.pm -- Changed the algorithm
to test against low and high value of the rule.

Identify, document and fix following problems if required:

- How frequently rules information is loaded in to the memory?

- Change data structure used within Rulebase.pm, and clean the
code.

- One of Low and high value within "constant" rule type can be
optional.

- SNMP probes --- do not send information with "zero" value
if SNMP session object creation is successful, but does not
return any value because of time-out.

- Use 100 seconds instead of 300 seconds for information like
memfree, loadavg parameters.

- Add a utility to periodically purge old data from reported_events.

How much days old data to retain must be stored in "globalparams"
table.

- Graph.pm -- "cascade*" and "draw*" functions are almost identical,
with lots of repeated code. Can it be identified and cleaned up?

- For "no change" type of rule, comparison must be done by comparing
against the +/- 1% range, instead of comparing for equality.

- What happens if specific event is not arrived but the last value
arrived represents the exception?

- How to find when last the particular event had arrived? Is it
important to display it somewhere? Shall an alert be generated
if particular event does not arrive for some time, or no events
arrive from a specific host for some time?

- Check Weekly graph X axis. One extra tick appears between first
and second day.

- Permission and ownership for /opt/merce/doc/graphs/*_LAST*

Ownership: merce.merce
Permission: -rw-------

Permission merce.merce is required for "utime" function call,
which changes the time of these files within "process_event"
script. "utime" function is called as

"utime currtime, currtime, filename"

If files are not owned by "merce", e.g. same permission as which
"process_event" runs, "utime" function fails. If "utime" was used
with "utime undef, undef, filename", it would have succeeded if
file permissions and ownership are like

Ownership: bin.merce
Permission: ----rw----

Decide whether to change the code, and depending on it, the ownership
and permission of the file.

- Moved /opt/merce/lib/Agentdconfig to /etc/merce/Agentdconfig.

Under high-availability mode with two active servers, e.g. one running
mail and another another one file server, /opt/merce/ is synchronized
and so do /opt/merce/lib/Agentdconfig, which is not desirable.

-x-

30 March 2007

- Fixed one bug in drawweekly function of Graphs.pm. An extra tick was
visible on X axis.

- At AZB, reported_events tables had crashed. Way to repair the crashed
table is:

repair table reported_events

Because of this reason, accept_event had failed, and process_event had
not generated a graph.

- Rule base: Bugs to be fixed for average change, and average rate of
change. The lower value for "ts" in reported_events, it should be
not lower than double the period mentioned in the note or 24 hours,
whichever is lower.

- Add a new agent -- nusers -- to measure number of users under various
categories, say once in a day.

- mex: Now mode 2 accepts up to 10 arguments. Integrate mex with agentd
and verify that "mex" enabled events are working.

- Averaging has a bug for weekly, monthly and yearly data, it seems.
This need to be sorted out.

-x-

14 May 2007

For SLES 10, stock perl-SNMP module crashes.

For 32 bit systems, please use
rpm -U net-snmp-5.3.0.1-25.5.i586.rpm perl-SNMP-5.3.0.1-25.5.i586.rpm

For 64 bit systems, please use
rpm -U net-snmp-5.3.0.1-25.5.x86_64.rpm perl-SNMP-5.3.0.1-25.5.x86_64.rpm

New Mex can be used to call scripts (agents) within /opt/merce/lib/secure.
Modified all instances of "runsetuidscripts" within "agentd" to
"mex". Now, "nsuid", "qmsgsendmail", "qmsgot*" agents are tested.

Modified permissions of scripts under /opt/merce/lib/merce/ to "0550"
to satisfy needs of "mex". Ideally, "mex" should be changed to check
for "0050" instead of "0550".

Fixed bugs within "listqueuefiles", "listuucpqueue" and "grepmqueue".
These scripts are under test.

Within "agentd", review "du*" agents. They will require modifications
to check "du" for any directory, not just one owned by "merce". "mex"
have to be used for this.

Pre-create directories for Merce Insight agents within
/opt/merce/doc/graphs/<agentname> with permissions
of 0070 and ownership of "bin.merce". This should be done
at installation time.

-x-