0

We have a 2-node cluster configured using red hat high availability (pacemaker) but without shared storage i.e. each of the node has its own disk /dev/sda and /dev/sdb.

We are not allowed to configure a shared LUN (icsi) for the two nodes to have access to. NFS is also not an option.

Is it safe to format /dev/sdb as gfs2 on node 1 and mount /dev/sdb on node 2 so that if applications writing to both nodes will not corrupt data in /dev/sdb? If not what's the alternate way to have an active-active solution e.g. through vmware sharing disk if possible maybe?

Note that the 2 nodes are provisioned as virtual machines in vmware.

--Edit: Using cephfs or glusterfs would require us to configure a separate cluster of at least three nodes which would take us more resources.

Linzynarcon
  • 1
  • 1
  • Does this answer your question? [How to replicate existing files using GlusterFS](https://serverfault.com/questions/575676/how-to-replicate-existing-files-using-glusterfs) – djdomi Aug 13 '22 at 17:56
  • Why do you have so many limitations? Is this a political issue that just requires buy-in and better justification to do it the correct way? – ewwhite Aug 13 '22 at 20:49
  • Do you have shared storage connected to ESXi hosts. Maybe vSAN. Your VMs should have some kind of shared storage to build a red hat cluster. https://kb.vmware.com/s/article/2151774 – Stuka Sep 08 '22 at 19:33

1 Answers1

0

I dont know what kind of solution are you running on your servers, but for example on Oracle RAC linux VM can share a virtual disk enabling multi-writer. With this option you avoid to use "shared storage" like iscsi or nfs. If you use this solution check that you software is supported with this kind of configuration.

Juan Herrera
  • 31
  • 2