<table cellspacing="0" cellpadding="0" border="0"><tr><td valign="top">The clustering will only works if you run same operating systems on top of same hardware platform ppc, intel!<a href="https://overview.mail.yahoo.com?.src=iOS"><br/><br/>Sent from Yahoo Mail for iPhone</a></td></tr></table>            <div id="_origMsg_">
                <div>
                    <br />
                    <div>
                        <div style="font-size:0.9em">
                            <hr size="1">
                            <b>
                                <span style="font-weight:bold">From:</span>
                            </b>
                            Digimer <lists@alteeve.ca>;                            <br>
                            <b>
                                <span style="font-weight:bold">To:</span>
                            </b>
                            linux clustering <linux-cluster@redhat.com>;                                                                             <br>
                            <b>
                                <span style="font-weight:bold">Subject:</span>
                            </b>
                            Re: [Linux-cluster] 2-node cluster fence loop                            <br>
                            <b>
                                <span style="font-weight:bold">Sent:</span>
                            </b>
                            Thu, Jun 12, 2014 4:25:14 PM                            <br>
                        </div>
                            <br>
                            <table cellspacing="0" cellpadding="0" border="0">
                                <tbody>
                                    <tr>
                                        <td valign="top">Have you tried simple things like disabling iptables or selinux, just to <br clear="none">test? If that doesn't work, and it's a small cluster, try unicast and <br clear="none">see if that helps (again, even if just to test).<br clear="none"><br clear="none">On 12/06/14 10:29 AM, Arun G Nair wrote:<br clear="none">> We have multicast enabled on the switch. I've also tried the<br clear="none">> multicast.py tool from RH's knowledge base to test multicast and I see<br clear="none">> the expected output, though the tool uses a different multicast IP(<br clear="none">> guess that shouldn't matter). I've tried increasing the post_join_delay<br clear="none">> to 360 seconds to give me enough time to check everything on both the<br clear="none">> nodes. One node still gets fenced. `clustat` output says the other node<br clear="none">> is offline on both
 servers. So one node can't see the other one ? This<br clear="none">> again points to issue with multicast. Any other clues as to what/where<br clear="none">> to look ?<br clear="none">><br clear="none">><br clear="none">> On Wed, Jun 11, 2014 at 8:33 PM, Digimer <<a shape="rect" ymailto="mailto:lists@alteeve.ca" href="javascript:return">lists@alteeve.ca</a><br clear="none">> <mailto:<a shape="rect" ymailto="mailto:lists@alteeve.ca" href="javascript:return">lists@alteeve.ca</a>>> wrote:<br clear="none">><br clear="none">>     On 11/06/14 10:48 AM, Arun G Nair wrote:<br clear="none">><br clear="none">>         Hello,<br clear="none">><br clear="none">>              What are the reasons for fence loops when only cman is<br clear="none">>         started ? We<br clear="none">>         have an RHEL 6.5 2-node cluster which goes in to a fence loop<br clear="none">>        
 and every<br clear="none">>         time we start cman on both nodes. Either one fences the other.<br clear="none">>         Multicast<br clear="none">>         seems to be working properly. My understanding is that without<br clear="none">>         rgmanager<br clear="none">>         running there won't be a multicast group subscription ? I don't<br clear="none">>         see the<br clear="none">>         multicast address in 'netstat -g' unless rgmanager is running. I've<br clear="none">>         tried to increase the fence post_join_delay but one of the nodes<br clear="none">>         still<br clear="none">>         gets fenced.<br clear="none">><br clear="none">>         The cluster works fine if we use unicast UDP.<br clear="none">><br clear="none">>         Thanks,<br clear="none">><br clear="none">><br clear="none">>     Hi,<br
 clear="none">><br clear="none">>        When cman starts, it waits post_join_delay seconds for the peer<br clear="none">>     to connect. If, after that time expires (6 seconds by default,<br clear="none">>     iirc), it gives up and calls a fence against the peer to put it into<br clear="none">>     a known state.<br clear="none">><br clear="none">>        Corosync is what determines membership, and it is started by<br clear="none">>     cman. The rgmanager only handles resource<br clear="none">>     start/stop/relocate/recovery and has nothing to do with fencing<br clear="none">>     directly. Corosync is what uses multicast.<br clear="none">><br clear="none">>        So as you seem to have already surmised, multicast is probably<br clear="none">>     not working in your environment. Have you enabled multicast traffic<br clear="none">>     on the firewall? Do your switches support
 multicast properly?<br clear="none">><br clear="none">>     digimer<br clear="none">><br clear="none">>     --<br clear="none">>     Digimer<br clear="none">>     Papers and Projects: <a shape="rect" href="https://alteeve.ca/w/" target="_blank">https://alteeve.ca/w/</a><br clear="none">>     What if the cure for cancer is trapped in the mind of a person<br clear="none">>     without access to education?<br clear="none">><br clear="none">>     --<br clear="none">>     Linux-cluster mailing list<br clear="none">>     <a shape="rect" ymailto="mailto:Linux-cluster@redhat.com" href="javascript:return">Linux-cluster@redhat.com</a> <mailto:<a shape="rect" ymailto="mailto:Linux-cluster@redhat.com" href="javascript:return">Linux-cluster@redhat.com</a>><br clear="none">>     <a shape="rect" href="https://www.redhat.com/__mailman/listinfo/linux-cluster"
 target="_blank">https://www.redhat.com/__mailman/listinfo/linux-cluster</a><br clear="none">>     <<a shape="rect" href="https://www.redhat.com/mailman/listinfo/linux-cluster" target="_blank">https://www.redhat.com/mailman/listinfo/linux-cluster</a>><br clear="none">><br clear="none">><br clear="none">><br clear="none">><br clear="none">> --<br clear="none">> Arun G Nair<br clear="none">> Sr. Sysadmin<br clear="none">> Dimension Data | Ph: (800) 664-9973<br clear="none">> Feedback? We're listening <<a shape="rect" href="http://www.surveymonkey.com/s/XRCYXBH" target="_blank">http://www.surveymonkey.com/s/XRCYXBH</a>><div class="yqt5833352895" id="yqtfd35959"><br clear="none">><br clear="none">><br clear="none"><br clear="none"><br clear="none">-- <br clear="none">Digimer<br clear="none">Papers and Projects: <a shape="rect" href="https://alteeve.ca/w/" target="_blank">https://alteeve.ca/w/</a><br
 clear="none">What if the cure for cancer is trapped in the mind of a person without <br clear="none">access to education?<br clear="none"><br clear="none">-- <br clear="none">Linux-cluster mailing list<br clear="none"><a shape="rect" ymailto="mailto:Linux-cluster@redhat.com" href="javascript:return">Linux-cluster@redhat.com</a><br clear="none"><a shape="rect" href="https://www.redhat.com/mailman/listinfo/linux-cluster" target="_blank">https://www.redhat.com/mailman/listinfo/linux-cluster</a><br clear="none"></div></td>
                                    </tr>
                                </tbody>
                            </table>
                    </div>
                </div>
            </div>