On Mon, Aug 10, 2009 at 1:33 AM, Trevor Benedict<[email protected]> wrote: > > Does anyone know of a hardware raid system that would have (in the same box) > 2 raid cards, and it act like a RAID 1 or 1+0 or even a RAID 5/6? Same > mobo/cpu. > > I would think someone like 3ware would have such a thing. This way you don't > have to have a 2nd xU mirror, and have to do "real time" replication, and IP > stuff.
I've never messed with any hardware that does this, but it seems awful difficult to me to try and implement that. I assume you're already using the model that's 12 or 16 drives and just want to add more? If you're using the 4 or 8 drive model, my uninformed advice would be to just upgrade to a larger card rather than try and link two together. I can't even imagine how much it would slow down the system for the kernel to have to track which PCI device it would have to talk to for each chunk of disk IO required. Or are you saying that the kernel really only talks to one and then that one distributes the data over some external bus connection that is direct from card to card? That puts the bottleneck on the card to pci-x bus connection, instead of spreading out the data writes/reads across two separate pci-x bus connectors. I may be misunderstanding something. > I only bring this up because I had a 3ware card fail on me last Monday. It > was a MySQL master. Which 3Ware card? We've got a box that's used as a central store for mysqldumps (or innodbbackup, depending on the system) with two 3Ware 9650SXU-8LP cards, both in RAID 5 with 1 TB drives (931 GB drives for the true 1024 believers), so the first controller has 4.55 TB, and the second has 5.46TB. The second controller just had an issue where one drive failed, and the spare did not automatically get added to the array and start a rebuild. Problem was, we didn't have it marked as a spare. PEBKAC. damn... BTW, if you are not using the 3dm2 (aka tdm2) app which gives you a web based (appears to be written in java) frontend for the tw_cli management tool, you are missing out. It is seriously very useful and seems to give a better picture of what is happening when there are problems, and gives you facilities to automate email to you/sysadmins when there is an issue with the controller or drives. -- Regards... Todd
