[Techtalk] Drive Failures

Magni Onsoien magnio+lc-techtalk at pvv.ntnu.no
Fri May 28 20:38:59 EST 2004


On 2004-05-28 08:31:19 +0100, Telsa Gwynne said:

> The only possible thing I can think of is to ask whether all this
> succession of drives were originally bought as a batch from the
> same place. 
> 
> I was amazed to find both my RAID discs dying within hours of each
> other. This became less amazing when I found that the serial numbers
> were separated by two digits: they were from the same bad batch or
> something. Apparently this sort of thing happening is well-known to 
> everyone except me. 

We had a test system in our office a few of years ago. Two days
before we were ready to ship it for production a disk in one of the
clients failed. We called the vendor and they said "ah, yes, it's a
faulty series of disks, we will send a technician to replace it."

So they actually KNEW this problem, but didn't bother to notify their
clients - it was probably too little possibility that the disk would
actually fail so they didn't want to replace all disks. Luckily the
techie came within an hour, replaced the disk and checked if the other
disks were in the same faulty serie. Which they were. We went rather
furious (at least inside fume!) and he promised to get us a pile of new
disks by next day. This time the disks were from four or five different
series (in total) and we distributed them so that no more than one or
two were from the same serie in the RAIDs (RAID5).

And we have had no other disk faults after that :-)

(We also notified a competitor of us, since we happened to know they
were putting a huge db system for a big client into production a few
days later. They were really glad for that - their disks were also from
the same serie, they just had five times as many of them and their
client was a more demanding one.)


Magni :)
-- 
sash is very good for you.


More information about the Techtalk mailing list