Re: [Linux-cluster] question about number of fencing devices needed for a two node cluster

Many thanks for the advice, Digimer. regards. > I know it would be desirable to have two devices for a fully redundant > configuration, but after reading some examples from the docs (they are > meant for two power switch), I still cannot understand why a single > power switch connected to both s

Re: [Linux-cluster] question about number of fencing devices needed for a two node cluster

Thanks for the tip, Jakov. regards, alvaro > ¿any experiences over this issue? It's because you still have SPOF. In this case, SPOF is the electronic module of the powerswitch, so, if the electronics go down, there's no way to fence the node. It would be better to have for example iDRAC or IPMI

Re: [Linux-cluster] question about number of fencing devices needed for a two node cluster

On 12/06/2010 10:11 PM, Alvaro Jose Fernandez wrote: > Hi, > > > > I would like to know about wheter it would suffice for a two-node RHCS > cluster a single power switch (APC 7921) fencing device. The power > switch has 8 power outlets and I intend to use four of them for each > node's dual pow

Re: [Linux-cluster] question about number of fencing devices needed for a two node cluster

On 12/06/2010 04:11 PM, Alvaro Jose Fernandez wrote: Hi, I would like to know about wheter it would suffice for a two-node RHCS cluster a single power switch (APC 7921) fencing device. The power switch has 8 power outlets and I intend to use four of them for each node's dual power supplies. I k

[Linux-cluster] question about number of fencing devices needed for a two node cluster

Hi, I would like to know about wheter it would suffice for a two-node RHCS cluster a single power switch (APC 7921) fencing device. The power switch has 8 power outlets and I intend to use four of them for each node's dual power supplies. I know it would be desirable to have two devices

[Linux-cluster] How do I implement an unmount only filesystem resource agent

Hi, I am configuring a service that uses HA-LVM and XFS filesystem on top of it. The filesystem will be backed up by a separate script run from cron(8) creating an LVM snapshot of the filesystem and mounting it on a mountpoint. To have a foolproof HA service I need to: - Check, if the sna

Re: [Linux-cluster] Difference between -d and -s options of clusvcadm

If you "disable" a service it won't start up again without manual intervention, like with "clusvcadm -e...". If you "stop" a service and let's say the node it was running on was rebooted, your service will start up on another node in the cluster if it was configured to do so. Greg Charles gch