[BBLISA] Moving from RAID 0 to LVM RAID?

[Edward Ned Harvey]

> > #1  If a disk goes bad, I expect hotswappable drives, with a red
> > blinking light.
> 
> Why would this require HW RAID?

Have you ever seen a chassis with hotswappable drives, that wasn't
preconfigured with a hardware raid card?  In theory, there's no reason
hotswappable drives & illuminated "disk failure" lights would require HW
raid, but I know I've never seen it without the HW raid card.


> Yes, if you cache writes to RAM, you potentially
> lose data, which is why no one does this. ;)
> 
> It's also why you don't want to enable write caching on most HW RAID
> controllers -- one power outage and oops, corrupt array.  The 3ware
> folks, at least, document this, BTW.
> 
> The good HW RAID controllers, and proper implementations of SW RAID,
> use something like NVRAM to solve this issue.  Power outage, kernel
> panic, whatever, the log of writes/requests/etc is stored in a battery-
> backed area so that at the next boot the system can replay the log and
> your array/filesystems/etc are safe -- writes only get acked once the
> information is stored in NVRAM.  (the newer 3ware cards, BTW, have a
> battery backup unit (BBU) which you can attach to the card, thereby
> making the write cache RAM safe to use)
> 
> 
> SW RAID gives you some benefits you don't get in most HW RAID setups,
> BTW, such as handling data integrity issues, and specifically data
> corruption on/from disk.  Implementations from NetApp, Sun (ZFS), etc,
> do SW RAID w/ block checksums, which let you detect when a block on
> disk is not what it should be.  Since it's part of a RAID setup, it
> lets you attempt to correct the data otherwise throwing an error
> instead of just returning invalid data.

I'm trying to understand what you're talking about here.  

On the one hand, you say "you don't want to enable write caching on HW raid"
and "nobody does this," but then on the other hand you also say "attach the
BBU to enable it" and seem to be in favor of SW raid.  The HW write cache is
enabled by default on dell PERC, which ships with the BBU installed, and
I've done the same with 3ware cards.  The performance benefit in the most
extreme cases is 60x faster, but typically 4x to 5x faster with the cache
enabled.  My answer is - just ensure you have your BBU and always ensure
your write cache is enabled.

My point is, this is something you can do with HW raid, which you can't do
with SW raid.  Unless you were insane and found some way to enable write
caching in the kernel.

Based on discussions here, it sounds to me like the best overall solution
would be:

HW raid card, with cache enabled, but no raid set configured.  Let the HW
perform caching acceleration, but present JBOD to the OS, and then let the
OS (presumably solaris in this case) use ZFS to perform
performance/reliability benefits of multiple disks.