Hi Kerry Based on your previous email I'm somewhat confused what you're actually trying to do. It sounds like you want to have 3 iSCSI servers behind two "gateway" type servers. The problem with this is unless you format the iSCSI volumes with a cluster filesystem then you can't mount the same block device on the two "gateway" servers and there really isn't anything that can be done to make that situation work. Since you're mounting the iSCSI volumes onto the Openfiler boxes and OF doesn't have any native support of cluster file systems as far as I know, there would be patching the kernel at that stage at the very least.
I think the problem is you're thinking of your storage architecture like you would think of a web farm. What it sounds like you want to do is protect against one of the gateway servers failing but that doesn't protect against one of the rear iSCSI servers from failing (perhaps you were planning to run software raid over the iSCSI volumes, not sure how that would work but it should be possible although it still doesn't remove the cluster filesystem requirement). I think the best you could do would be to setup multiple HA configurations so for each active OF box you have a secondary in passive mode. If the active fails then the passive takes over the IP address etc and everything continues to work, weather these boxes are straight NAS or iSCSI mounted on the windows servers it shouldn't matter. Since you've got an odd number of boxes this obviously won't be a perfect solution but I think it's the best you could do. One other possibility that _may_ work is to have the passive gateway mount the iSCSI volume read-only and if it loses the heartbeat from the active node then remount it RW and have it takeover the IP etc. Doing this, and even doing what I think you were suggesting in your previous email, you wouldn't actually need any replication at all as you're using the same storage devices, it's effectively a SAN and a shared storage pool, all you would need is the heartbeat setup properly. The risk here is if the failover doesn't happen fast and clean and there are still connected clients to the gateway that's just "failed" that write to the storage you'll likely hose the filesystem. Anyway, more info is required before knowing exactly how you could make it work Dave From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Kerry L. Kriegel Sent: Tuesday, 19 February 2008 6:56 a.m. To: [email protected] Subject: [OF-users] 2.3 HA question OpenFilers ---- I understand that the OF 2.2 HA was a single "active" node with replication to the "standby" node. I have installed OF 2.3 and am about to implement HA using the instructions for 2.2 (the only ones I can find). My question is.....Is OF 2.3 HA capable of performing with BOTH nodes as "active" for load balancing and fault tolerance? If so, are there any instructions out there for OF 2.3 HA? Thank you, Kerry L. Kriegel
_______________________________________________ Openfiler-users mailing list [email protected] https://lists.openfiler.com/mailman/listinfo/openfiler-users
