[Pacemaker] pingd problems

[Dalibor Dukic]

On Sat, 2010-06-05 at 15:36 +0200, Dalibor Dukic wrote:
> I have problem with ping RA not correctly updating CIB with appropriate
> attributes when doing fresh start. So afterwards IPaddr2 resources wont
> start. 

Have anyone had chance to get a peek at this? 

My setup consists from two nodes doing ACTIVE/ACTIVE solution for IVR
yate service (heartbeat with pacemaker 1.0.8). 

I have ping monitor to default gateway which is cloned between nodes and
every node have self floating address (VIP1 and VIP2). 

If one node fails or ping monitor fails other node takes floating
address. 

primitive L3_ping ocf:pacemaker:ping \
	params host_list="10.63.97.25" multiplier="100" \
	op monitor interval="10s" timeout="5s" on-fail="standby"
clone L3_ping_clone L3_ping \
	meta globally-unique="false" target-role="Started"
primitive VIP1 ocf:heartbeat:IPaddr2 \
	params ip="10.63.97.28" \
	op monitor interval="15s" 
primitive VIP2 ocf:heartbeat:IPaddr2 \
	params ip="10.63.97.29" \
	op monitor interval="15s" 

Also yate LSB service is cloned between nodes.

primitive Yate lsb:yate \
	op monitor on-fail="standby" interval="15s"
clone Yate_clone Yate \
	meta globally-unique="false" target-role="Started"


So, problem lies in location constraints:

location LOC_VIP1 VIP1 \
	rule $id="LOC_VIP1-rule" 100: #uname eq 7it-ivr-1
location LOC_VIP1_CONNECTED VIP1 \
	rule $id="LOC_VIP1_CONNECTED-rule" -inf: not_defined L3_ping or L3_ping
number:lte 0
location LOC_VIP2 VIP2 \
	rule $id="LOC_VIP2-rule" 100: #uname eq 7it-ivr-2
location LOC_VIP2_CONNECTED VIP2 \
	rule $id="LOC_VIP2_CONNECTED-rule" -inf: not_defined L3_ping or L3_ping
number:lte 0


After configuring location constraint to tell the cluster to only run
the floating address on a node with a working network connection to the
default gateway my IPaddr resource won't even start on every cluster
node. 

Node 7it-ivr-1 (5f783be2-eff1-4db7-9b94-0b13a4670bb4): online
        Yate:0  (lsb:yate) Started
        L3_ping:0       (ocf::pacemaker:ping) Started
Node 7it-ivr-2 (e8b2c3be-1d32-43d8-9876-d73642693ccf): online
        L3_ping:1       (ocf::pacemaker:ping) Started
        Yate:1  (lsb:yate) Started


Current allocation scores:

root at 7it-ivr-1:~# ptest -sL
Allocation scores:
native_color: VIP2 allocation score on 7it-ivr-1: -1000000
native_color: VIP2 allocation score on 7it-ivr-2: -1000000
clone_color: Yate_clone allocation score on 7it-ivr-1: 0
clone_color: Yate_clone allocation score on 7it-ivr-2: 0
clone_color: Yate:0 allocation score on 7it-ivr-1: 1
clone_color: Yate:0 allocation score on 7it-ivr-2: 0
clone_color: Yate:1 allocation score on 7it-ivr-1: 0
clone_color: Yate:1 allocation score on 7it-ivr-2: 1
native_color: Yate:0 allocation score on 7it-ivr-1: 1
native_color: Yate:0 allocation score on 7it-ivr-2: 0
native_color: Yate:1 allocation score on 7it-ivr-1: -1000000
native_color: Yate:1 allocation score on 7it-ivr-2: 1
native_color: VIP1 allocation score on 7it-ivr-1: -1000000
native_color: VIP1 allocation score on 7it-ivr-2: -1000000
clone_color: L3_ping_clone allocation score on 7it-ivr-1: 0
clone_color: L3_ping_clone allocation score on 7it-ivr-2: 0
clone_color: L3_ping:0 allocation score on 7it-ivr-1: 1
clone_color: L3_ping:0 allocation score on 7it-ivr-2: 0
clone_color: L3_ping:1 allocation score on 7it-ivr-1: 0
clone_color: L3_ping:1 allocation score on 7it-ivr-2: 1
native_color: L3_ping:0 allocation score on 7it-ivr-1: 1
native_color: L3_ping:0 allocation score on 7it-ivr-2: 0
native_color: L3_ping:1 allocation score on 7it-ivr-1: -1000000
native_color: L3_ping:1 allocation score on 7it-ivr-2: 1


>From this output I can see that VIP1 and VIP2 resource is not started
because of wrong scores. 

native_color: VIP1 allocation score on 7it-ivr-1: -1000000
native_color: VIP2 allocation score on 7it-ivr-2: -1000000

It is very annoying to have cluster but can't use ping monitor to test
default gateway reachability. 

I would kindly appreciate if someone could help me with resolving
described problem.

best regards, Dalibor