[Linux-HA] crm, pingd, bond - cause with migrate resources

[Pavel Sebek]

Since already moderator my mail not approval because was bigger than
100k I send again with gzip's attachement.

Hi,

i could this situation try onlz todaz because cluster is running at
production already.

In attachment are output cibadmin -Q (cib.xml) before lose connection to
the ping node and ouput after change state ping node on dead
(cib_ping_node_dead.xml). I put also logs. Lose conection to the ping
node i made by iptables (iptables.txt).
I expected after what node1 (cn42x1) losed connection to the ping node
that the all resources will move to the second node (cn42x2).
But the resources were left on the node which losed connection to the
ping node.

Pavel

PS:Is there posibility how found out node's score?

Andrew Beekhof napsal(a):

> On 8/28/06, Pavel Sebek <sebek at via.cz> wrote:
>
>> Hi,
>>
>> I tried '<nvpair id="pingd-multiplier" name="multiplier"
>> value="INFINITY"/>' without success.
>>
>> Now I tried another:
>> score for group resource on node1 is 200
>> score for group resource on node2 is 199
>> resource_stickiness is 155
>> multiplier for pingd is 180
>> one ping node
>>
>> I think that after start both nodes the all resources are running on
>> node1. It's true. IMHO node1's score is (200 + 180 + 155) = 535 (score
>> of node + pingd + resource_stickiness)
>> And node2's score is (199+180) = 379
>> I think that after lose link on interface at node1 the node1's  score
>> would be (200 + 155) = 355 and node2's score would be (199 + 180) = 379
>> and therefore the all resources would moves to the node2.
>>
>> But It doesn't work ;-(
>>
>> Why?
>
>
> If you get it into this situation again (where it should be on the
> "other" node because of pingd, can you run "cibadmin -Q" and attach
> the result here?
>


-------------- next part --------------
A non-text attachment was scrubbed...
Name: cib_ping_node_dead.xml.gz
Type: application/gzip
Size: 3790 bytes
Desc: not available
Url : http://lists.community.tummy.com/pipermail/linux-ha/attachments/20060905/7dd05137/cib_ping_node_dead.xml.bin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ha-debug.gz
Type: application/gzip
Size: 1895 bytes
Desc: not available
Url : http://lists.community.tummy.com/pipermail/linux-ha/attachments/20060905/7dd05137/ha-debug.bin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ha-log.gz
Type: application/gzip
Size: 1845 bytes
Desc: not available
Url : http://lists.community.tummy.com/pipermail/linux-ha/attachments/20060905/7dd05137/ha-log.bin
-------------- next part --------------
[root at xxxx ha]# iptables -vnxL
Chain INPUT (policy ACCEPT 5120674 packets, 1039892387 bytes)
    pkts      bytes target     prot opt in     out     source               destination

Chain FORWARD (policy ACCEPT 0 packets, 0 bytes)
    pkts      bytes target     prot opt in     out     source               destination

Chain OUTPUT (policy ACCEPT 2989188 packets, 446932188 bytes)
    pkts      bytes target     prot opt in     out     source               destination
[root at xxxx ha]# /usr/local/bin/stop_ping-node
[root at xxxx ha]# iptables -vnxL
Chain INPUT (policy ACCEPT 5120843 packets, 1039919976 bytes)
    pkts      bytes target     prot opt in     out     source               destination

Chain FORWARD (policy ACCEPT 0 packets, 0 bytes)
    pkts      bytes target     prot opt in     out     source               destination

Chain OUTPUT (policy ACCEPT 2989282 packets, 446945668 bytes)
    pkts      bytes target     prot opt in     out     source               destination
       1      137 DROP       icmp --  *      *       0.0.0.0/0            10.6.35.1
[root at xxxx ha]