AMF goes to maintenance state

Description

Error Message

[SNIP from /var/adm/messages]
Nov 21 14:55:19 apg-sol10 svc.startd[7]: [ID 652011 daemon.warning] svc:/system/amf:default: Method "/lib/svc/method/amf stop" failed with exit status 1.
Nov 21 14:55:19 apg-sol10 last message repeated 2 times
Nov 21 14:55:19 apg-sol10 svc.startd[7]: [ID 748625 daemon.error] system/amf:default failed: transitioned to maintenance (see 'svcs -xv' for details)

[SNIP from the output of "svcs -xv"]
svc:/system/amf:default (Veritas Asynchronous Monitoring Framework (AMF) Init service)
State: maintenance since Mon Nov 21 14:55:19 2011
Reason: Method failed repeatedly.
See: http://sun.com/msg/SMF-8000-8Q
See: amfconfig(1M)
See: amfstat(1M)
See: /var/svc/log/system-amf:default.log
Impact: This service is not running.

[SNIP from "/var/svc/log/system-amf:default.log"]
[ 11\267\356 21 14:55:19 Executing stop method ("/lib/svc/method/amf stop") ]
Stopping AMF...
AMF amfconfig ERROR V-292-2-405 AMF_UNCONFIG failed, return value = -1 AMF: Unable to unload module
[ 11\267\356 21 14:55:19 Method "stop" exited with status 1 ]

Cause

Traditional monitoring of resources means that the VCS Agent will periodically run monitor routines for each resource.
Intelligent Monitoring Framework (IMF) monitoring means that the VCS Agent will rely on the VCS IMF subsystem to get resource state change notifications so that periodic monitoring is minimized.

To stop IMF monitoring for resources, it is sufficient to change IMF mode to 0. This will NOT unload the AMF driver but will stop IMF monitoring of resources. At this stage, any attempt to stop AMF driver will fail.

Resolution

The customer might want to completely unload AMF driver from the system. In that case, he can use the following steps:

(A) Elegant way if Agent down time is acceptable

a. Stop all the agents registered with AMF kernel driver. The "Registered Reapers" section in the amfstat output indicates which all agents are registered with AMF.

# haagent -stop -force -sys

b. Stop AMF

b-1. Stop the AMF kernel driver using the platform specific command:

- On Solaris 9:

# /etc/init.d/amf stop

- On Solaris 10

# svcadm disable amf

b-2. Set the value of the AMF_START variable to 0 in "/etc/default/amf"

b-3. Remove the AMF kernel driver using the following command:

# rem_drv amf

c. Start all these agents

# haagent -start -sys

(B) Force unconfig of AMF if Agent downtime is not acceptable

a. Replace the following line in "/lib/svc/method/amf" with "$AMFCONFIG -oUf" to unconfig AMF driver forcibily even if agents are still registered with it.

[SNIP from "/lib/svc/method/amf"]
.....
185 echo "Stopping AMF..."
186 if mod_isloaded ; then
187 $AMFCONFIG -oUf <<< Here