[Devmon] devmon.cfg best way to set it up

[Joshua K. <fo...@tr...>]

I know a lot of the people in this group are using devmon to poll a lot of
cisco equipment along with others.  I am trying to see what is the setup of
other people on there devmon.cfg.  I am using devmon to monitor 283 hosts
which are mostly all cisco.  Most of my devices are 2950's and 4948's but I
also have some 2970's, 3500's, 4500's, and 4 6509's, which are fully loaded.
The 6500's have probably 3-4 48 port gig blades and then 2-3 24-48 port sfp
ports.  And I notice devmon is exceeding cycle times, so I've up'd the cycle
time to 120 from 60.  I have changed the number of forks from 10-30 and
still having trouble exceeding cycle times.  I have recently converted over
my if_stat and if_load to use the 64 bit counters for all my switches.  And
I am noticing a new entry in my devmon.log "Timed out waiting for delta
transform on oid ifOutOps_T for gsonc-spg-6509-2 to complete" which is from
one of the if_load template for the 64 bit counters.

 

My server I am using is an IBM i346 series with 4 Intel(R) Xeon(TM) CPU
3.20GHz and 4-8 gig ram.  I use my server to run hobbit and devmon off of
it.  So I am trying to see what is the best way to setup my devmon.cfg to
allow me to poll at a descent rate and not use up the cpu and load.  I know
some people may be monitoring more or might have just tweaked there
devmon.cfg where it is better utilized.  As you can tell I am not using the
database function and if that is a better way to do it I will go that route.
Here is mine:

 

# Devmon config file

 

# Variable/value pairs in this file are delimited by an equals sign, and

# all values are case insensitive unless otherwise stated

 

######## LOCAL OPTIONS:

########

 

######## These options are local to this machine, and changing them will not

######## affect the runtime parameters of other nodes in a Devmon cluster

######## (assuming you are running in with MULTINODE=YES)

 

# MULTINODE: Determines if this node is a single node system. Acceptable

# values are 'yes' and 'no' [DEFAULT: NO]

MULTINODE=NO

 

# BBHOSTS: Location of the bb-hosts file on this system.  This file is only

# used when running devmon with the --readbbhosts flag.  In a multi-node

# cluster, the only system that needs to have this set is the one that

# the display server is running on;  other systems can leave this at
default.

# (case sensitive) [DEFAULT: /home/hobbit/server/etc/bb-hosts]

BBHOSTS=/opt/hobbit/server/etc/bb-hosts

 

# BBTAG:  This is the indentifier (aka tag) that you use in your bb-hosts

# which allows Devmon to determine which of the options after a host

# are Devmon specific.

# (case sensitive) [DEFAULT: DEVMON]

BBTAG=DEVMON

 

# SNMPCIDS: The SNMP cids used during the bb-hosts discovery process

# (i.e. running with the --readbbhosts flag) This should be a comma
delimited

# list of values;  you order the list from most common to least common, left

# to right, as this will help speed the device type discovery process.

# (case sensitive) [DEFAULT: public,private]

SNMPCIDS=xxxxxxx,xxxxxxxxx

 

# NODENAME: Sets the name of this Devmon node.  A value of of 'HOSTNAME'

# will cause the host to determine its hostname via the 'hostname' command

# (the non-FQDN name will be used).  Only useful if in multinode mode.

# (case sensitive) [DEFAULT: HOSTNAME]

NODENAME=xxxxxxxxxxxx

 

# PIDFILE: Sets the full path to the Devmon process ID file.  The directory

# containing this file should be read/writable by the user running the
Devmon

# process. (case sensitive) [DEFAULT: /var/run/devmon/devmon.pid]

PIDFILE=/var/run/devmon/devmon.pid

 

# LOGFILE: Sets the full path to the Devmon log file. The file should either

# exist and be writable by the Devmon user, or the Devmon user should have

# read/write privileges to the log directory (case sensitive)

# To disable logging to a file, leave this value blank (i.e. 'LOGFILE=')

# [DEFAULT: /var/log/devmon.log]

LOGFILE=/var/log/hobbit/devmon.log

 

# DBHOST: Should contain the IP or DNS name of the database server

# in a multinode cluster. [DEFAULT: localhost]

DBHOST=localhost

 

# DBNAME: Specifies the name of the Devmon database

# on the database server.  (case sensitive) [DEFAULT: devmon]

DBNAME=devmon

 

# DBUSER: Sets the name of the devmon user on the database.  This user
should

# have full read/write priveledges on the database specified by the DBNAME

# variable. (case sensitive) [DEFAULT: devmon]

DBUSER=devmon

 

# DBPASS: Sets the password used the DBUSER to connect to the DBNAME
database

# on DBHOST. (case sensitive) [DEFAULT: devmon]

DBPASS=devmon

 

 

 

 

 

######### GLOBAL OPTIONS:

#########

 

######### Changing any of these options (followed by running devmon with the

######### --syncconfig flag, if you are running in multinode mode) will
change

######### the global runtime parameters for all devmon nodes in this
cluster.

######### Change these options with care, as they can have a major impact on

######### way Devmon functions!

 

 

### BB/HOBBIT specific options

###

 

# BBTYPE: Determines how devmon will format its output pages.Accetable
values

# are 'bb' or 'hobbit', either each option setting the output format to that

# of its respective server type [DEFAULT: hobbit]

BBTYPE=hobbit

 

# DISPSERV: Should contain the DNS name or IP address of th BB or hobbit

# display server. [DEFAULT: localhost]

DISPSERV=localhost

 

# DISPPORT: Specifies the port on the remote display server to connect to

# [DEFAULT: 1984]

DISPPORT=1984

 

# MSGSIZE: Sets the max message size in bytes of the datagrams that Devmon

# sends to the bb display server [DEFAULT: 8096]

MSGSIZE=8096

 

# MSGSLEEP: The number of milliseconds to sleep between each message sent

# to the bb displayer server.  Set this number higher if your bb server

# is having problems coping with an influx of Devmon data. [DEFAULT: 10]

MSGSLEEP=10

 

 

### Polling control specific options

###

 

# CYCLETIME: The length of the polling cycle, aka how often Devmon

# will attempt to query remote hosts and send messages to the bb display

# server.  Measured in seconds.  [DEFAULT: 60]

CYCLETIME=120

 

# DEADTIME: The amount of time before the devmon cluster considers a

# particular node "dead".  This value should be at least twice your

# cycletime variable, if not a little bit more.  [DEFAULT: 180]

DEADTIME=240

 

# CLEARTIME: The amount of time in seconds Devmon will wait before

# reporting a clear test.  In high-latentcy networks, this helps

# prevent excessive amounts of clear reports. [DEFAULT: 180]

CLEARTIME=180

 

# NUMFORKS: The number of child processces that Devmon should spawn

# to handle SNMP transations.  Too many will bog down your machine,

# while too few will cause your test cycles to run much slower.

# It best to experiment with this, but the default works well.

# [DEFAULT: 10]

NUMFORKS=30

 

# NUMFORKS: The maximum amount of time in seconds that a child process

# should spend polling a single host.  The can help recover forks

# in case one hangs on a remote host.  This value should be greater

# than (SNMPTRIES x SNMPTIMEOUT) + 1 [DEFAULT: 30]

MAXPOLLTIME=30

 

### SNMP specific options

###

 

# SNMPTIMEOUT: Number of seconds before an snmp query will time out

# Setting this lower might increase the speed of your SNMP queries, but

# on high latentcy connections that might result in missed poll cycles

# [DEFAULT: 2] (must be greater than 0)

SNMPTIMEOUT=2

 

# SNMPTRIES: Number of attempts an SNMP query will make before it gives

# up.  More than 5 or so might be overkill here

# [DEFAULT: 5] (must be greater than 0)

SNMPTRIES=5