Is it possible use replicated storage in a network RAID 1 fashion?

Portal Home > Knowledgebase > Articles Database > Is it possible use replicated storage in a network RAID 1 fashion?

Posted by fourlightson, 10-25-2010, 11:53 AM
Hi, I was wondering if there was any way to replicate the functionality of the HP StorageWorks P4000 Virtual SAN Appliance using open source tools? Basically the HP VSA lets you replicate in the same way DRBD does but it lets you use both of the block devices in a RAID 1 fashion so you can read data from both the local storage and from the replicated storage on the other server. When I've tested DRBD I've used it in active/passive so if i have RAID 10 and DRBD sitting on top of it in both servers, one server does all the work while the other us just receiving the replicated block data in prep for the failover. I'd like to get test using both to read from at the same time, so if i had 30 XEN guests on one server the load would be spread over the storage on two or more physical servers. Hopefully that makes some sense. Any pointers would be most appreciated. thanks flo
Posted by YUPAPA, 10-25-2010, 01:23 PM
Have you looked at gluster? It lets you to read the data from either one server, however, the read and write are concentrated on one server and then replicate the changes to the other one. It is a file level replication rather than block replication, so you can easily read the files on either end. Also another idea I've thought of is to use raid over iscsi, but not sure about the performance on this one.
Posted by fourlightson, 10-25-2010, 01:40 PM
hi, yea, i looked at gluserfs but after some comments on various sites about it being slow i wasn't sure if it was the best way. I'll test it thought in vmware to see how it performs. i was also looking at old skool Network block devices and iscsi but never found a way to replicate/load balance the read/write using those. I'll test gluster. The HP VSA is a nice bit of kit but at £3000+ it's too pricey for us so i want to try and replicate is with Linux. thanks.
Posted by sectoroverload2k, 10-26-2010, 09:12 PM
there's a userspace application/module for linux called drbd (distributed redundant block device). There's a way that you can "cross configure" it to allow both nodes to write simultaneously. Here's the basic overview: Both server1 and server2 are running the linux-ha tools to enable nfs and drbd master/standby configurations Server1 is drbd master (/dev/drbd1 -> /shared/drbd) Server2 is dbrd standby (/dev/drbd1 -> /shared/drbd) server2 (standby) is nfs mounted to /shared/drbd on server1 in the event of a failure, linux-ha will swap the server roles. server1 (prev. master.. now standby) mounts to /shared/drbd on server2 and server2 becomes the drbd master. during this failover, there may be 2-3 seconds downtime. i had this configuration setup and working successfully for 2 years at my previous company.