Hi all Hi Brem and thanks for your answer. When I moved manually my service, it work then I thought my quorum was used. But it seems not in fact. Even if the case "has quorate" was checked in system-config-cluster. I have a special vlan dedicated for a heartbeat link between my two nodes. In fact, I have three vlan for my cluster, one for production (192.168.104.), one for admin (192.168.50.) and the last one for the heartbeat (192.168.30)(network only dedicated for my nodes). But you said : "Currently supported, only the clusternode (cluster.conf) corresponding IP addresses are used for heartbeating. Multiple heartbeat interfaces, referred as multiring is technicaly possiblebut not "yet" supported" Does it mean I can't define several different IP adress in my cluster.conf ? Then what about multicast ? Could I define there my "heartbeat IP" ? In reality, i "must" have an ip adress on each nodes dedicated for the heartbeat link... Another question : how to set directly in cluster.conf that my point of mount will have to use gfs2 and not gfs ? Because in system-config-cluster, gfs2 is not present. I have to modify cluster.conf with vi and propagate it manually. Luci knows gfs2. Sorry if I'm not clear ... :) mog Le samedi 06 mars 2010 à 00:14 +0100, brem belguebli a écrit : > On Fri, 2010-03-05 at 21:18 +0100, mogruith wrote: > > Hi all > > > > Today, my cluster crashed, then I have several questions to ask . > > > > > > - First of all, is there a kind of "heartbeat" on a quorum disk ? If > > yes, it means, I have two heartbeat on my cluster, one by the quorum, > > second one by a network link. Is it right ? > qdisk is definitely a heartbeat vector. Nodes regularly, at configured > interval value, write to a status block which modifies the timestamp. > cf man qdisk, chapter 2.1 Heartbeating and liveliness determination > > > > > - How to set the heartbeat link in my cluster conf ? Is the "monitor > > link" is a heartbeat when I set an IP adress to my service ? I have 2 > > cards just for the heartbeat network and I have two others cards for > > another vlans. > > > Currently supported, only the clusternode (cluster.conf) corresponding > IP addresses are used for heartbeating. > Multiple heartbeat interfaces, referred as multiring is technicaly > possiblebut not "yet" supported. > > - I use mkqdisk to configure my quorum disk. My two nodes were able to > > see it (I used mkqdisk -L on each node). My cluster was quorate. But > > when I killed my first node, the second one didn't take the service and > > the cluster told me it was inquorate ( Cluster is not quorate. Refusing > > connection. Error while processing connect: Connection refused). > > > If using qdisk in a 2 nodes cluster, two_nodes setting (cluster.conf) > must be =0. > Quorum depends on the number of votes you gave to each node and to the > quorum. > Cluster is quorate if number of active votes > 50% of the expected votes > (greater than) > mkqdisk is not sufficient, the qdisk device must also be defined in the > cluster.conf. > > Then service was down and cluster was strange. ccsd and qdiskd were > > still running. > > > > I relocated service manually and everything was right. Service moved to > > a node to another one without any problem. > > > How could you relocate the server manually if the cluster was not > quorate ? > > Well .. thanks for your help. > > > > mog > > > > -- > > Linux-cluster mailing list > > Linux-cluster@xxxxxxxxxx > > https://www.redhat.com/mailman/listinfo/linux-cluster > > > -- > Linux-cluster mailing list > Linux-cluster@xxxxxxxxxx > https://www.redhat.com/mailman/listinfo/linux-cluster -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
