[Vacm-develop] EMP deconfigure

[Dirk W. <di...@re...>]

Hi,

is there anyway to deconfigure EMP support for a node which
only has serial console access? the resulting GUI in hoover is
somewhat confusing, because all those nodes are listed as
dead.


the entries in vacm_configuration look like this:

#Definition for node 'my_node'
NODE:my_node
MODMAP:
VAR:EMP:SERIAL_PORT:NONE
VAR:EMP:PASSWORD:NONE
VAR:SERCON:BAUDRATE:9600
VAR:SERCON:DEVICE_ADDRESS:xxx.xx.xxx.xx-port
END_NODE

#Definition for node 'another_node'
NODE:another_node
MODMAP:
VAR:GLOBAL:IP_ADDRESS:xxx.xx.xx.xx
VAR:EMP:SERIAL_PORT:NONE
VAR:EMP:PASSWORD:NONE
VAR:SERCON:BAUDRATE:38400
VAR:SERCON:DEVICE_ADDRESS:/dev/ttyRxx
VAR:SYSSTAT:AUTH_PASSWORD:XXXXXX
VAR:RSH:AGENT:RSH
VAR:RSH:USER:root
END_NODE


all emp relevant entries were according to the docu created with
"ipc localhost EMP:CONFIGURATION:node_name:NONE:NONE".

another issue i worry about: i have only 4 machines speaking and configured
EMP, but 147 emp.loose processes! lsof'ing those emp.loose processes,
it appears to me like every process is only talking to my 4 emp
capable nodes, i only see /dev/ttyX connections to those machines
(and each ~500 pipes, they are a shared amongst the emp.loose processes).

pstree output is like:

-nexxus-+-apcups.loose---apcups.loose---apcups.loose---48*[apcups.loose]
        |-baytech.loose---baytech.loose---baytech.loose---48*[baytech.loose]
        |-emp.loose---emp.loose---emp.loose---144*[emp.loose]
        |-icmp_echo.loose
        |-msc.loose---msc.loose---msc.loose---48*[msc.loose]
        |-rsh.loose---rsh.loose---rsh.loose---48*[rsh.loose]
        |-sbt2.loose---sbt2.loose---sbt2.loose---48*[sbt2.loose]
        |-sercon.loose---sercon.loose---sercon.loose---48*[sercon.loose]
        |-sys_stat.loose
        |-va1000.loose
        `-vasenet.loose---vasenet.loose---vasenet.loose---48*[vasenet.loose]



if there's somebody still listening here, i would be happy if somebody
could tell me how to get rid of the confusing "dead" entries in hoover.
also i would like to know, how to restrict the number of "*-loose"
processes. i am afraid if i hook up the next 100 node cluster i am running
out of resources on my management node.



thx,
	~dirkw

-----------
Dirk Wetter
Renaissance Technologies/NY