[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: SATA SB700/SB80 io errors

On 3/9/2012 5:05 AM, Robert Gstoehl wrote:
> Hey there,
> Our DRBD primary machine expirenced a rather spontanous reboot some time ago.
> We were happily starting / stopping kvm virtual machines, syncing a
> new drbd resource and
> then this happened:

[snipped logs]

> node2 drbd primary
> HP Proliant Micro Server

> node3 drbd secondary
> HP Proliant Micro Server

/me tries not to laugh

> Samsung Spinpoint drives

/me tries harder not to laugh

You've taken two cheap college student dorm room 'servers', clustered
them, and sold the solution to a commercial client...

> Any help is appreciated to get this sorted out.

I have advice to offer, but you probably won't consider it helpful.  And
you'll likely find it offensive/insulting, as it is my intention to
publicly shame/humiliate you in to a proper way of thinking:

Don't sell your commercial clients $300 USD consumer desktops
masquerading as 'servers', and the cheapest consumer drives on the
planet for commercial use.  I still have trouble digesting the fact you
did such a thing.  There could have been many reasons for you doing so,
and no of them reflect positively on you, your company, or the way you
conduct business.

The only way to "fix" this, technically, is to drop a real SAS/SATA HBA
and 4 enterprise class SATA drives into each existing box.  This will
cost ~$2400 USD w/8x1TB drives.  The other option is to throw all the
junk hardware out and replace it with a single commercial quality box
with a quad core CPU, 8GB RAM, enterprise HBA or RAID card, and 4
enterprise SATA or SAS drives.  The cost is roughly the same as the
"upgrade solution above" using HBA+SATA, a couple hundred more for RAID,
and another couple hundred for 600GB 10k SAS drives.  You'll have 1.8TB
total RAID5 or 1.2TB less than the current system, but random disk IOPS
will be doubled vs SATA, which is desirable for VM workloads.


Reply to: