Greetings
In fact with DRBD also we can have GFS by making both primary. I have worked in such a setup. to further bolster fault tolerance, you can have Software RAID1 or 5 in each box. SATA drives are not expensive
Alternatively if you want more nodes to participate in the cluster, you can build a low cost iSCSI target using openfiler and give-em a handful of bonded NICs with jumbo frames to achieve decent performance. With this solution you will have a storage box outside. If you have rack mounting facility, try Sunfire x4150 it is just 1RU and a lot of thigs like 8 HDD bays, 4NICs Memory upto 64GB.
Hope that answers the question somewhat
Regards
Rajagopal
2008/8/25 Anuj Singh (अनुज) <anujhere@xxxxxxxxx>
sources.redhat.com/cluster/doc/nfscookbook.pdf
Now in the above howto, it is assumed that you are using SAN for storage
If budget constraints are there, One alterntive for two node cluster with shared storage achived with the storage box having two scsi channels and both the nodes having one scsi controller each.
else the following is applicable
AIUI your target is to use A & B (Cluster nodes) to provide high availability of nfs service to your client machine C.
a) If node A goes down B takes over the nfs service, provided to client C.
b) If node B goes down A takes over the nfs service. provided to client C.
c) At any particular time only A node or B node providing service to C.
c) Client C, should see the same data independent of A and B, whether A is providing nfs export or B.
sources.redhat.com/cluster/doc/nfscookbook.pdf
Now in the above howto, it is assumed that you are using SAN for storage
If budget constraints are there, One alterntive for two node cluster with shared storage achived with the storage box having two scsi channels and both the nodes having one scsi controller each.
else the following is applicable
Above thing can be achieved easily.
a) Get the DRBD (distributed redundant block device) and replicate data between A node, B node. RAID1
b) You need to create a virtual IP which floats between A node and B node along with startup of nfs service.
c) You client C will access the floating virtual IP.
above three things can be achieved by linux-heartbeat as well as RHCS, with RHCS you only have to manage DRBD primary. which can be handled with some small script.
In fact with DRBD also we can have GFS by making both primary. I have worked in such a setup. to further bolster fault tolerance, you can have Software RAID1 or 5 in each box. SATA drives are not expensive
Alternatively if you want more nodes to participate in the cluster, you can build a low cost iSCSI target using openfiler and give-em a handful of bonded NICs with jumbo frames to achieve decent performance. With this solution you will have a storage box outside. If you have rack mounting facility, try Sunfire x4150 it is just 1RU and a lot of thigs like 8 HDD bays, 4NICs Memory upto 64GB.
Hope that answers the question somewhat
Regards
Rajagopal
-- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
