[rescue] fire v210 upgrade

Thog of Cave People thog at thogscave.net
Wed Jul 4 13:41:20 CDT 2007


On Jul 4, 2007, at 7:13 AM, Carl R. Friend wrote:

>     I fully intend to look at more open systems, ones that can
> be properly upgraded, but since it's getting to the point where
> it's either Intel or SPARC my hand may be forced.  I am *not*
> going to gleefully join the Intel monoculture.

You also need to keep in mind the target audience.  It's not really  
hobbyists or small companies.  For example, I currently work at a  
shop that uses hundreds of SPARC systems, many of them V210s and  
V240s (racks full).  For us, quick replacement of failed components  
is of prime concern, and it's nice that the system board is a FRU.   
As for upgrades, again it's easier.

Another factor is the skill needed to replace components.  In the  
days when tolerances were less, it was O.K. to kick a VME board to  
seat it.  But, as a Senior Engineer for Sun (one of my past lives), I  
personally witnessed an SSE (who should know better) kick a board on  
an E10K to seat it.  I almost fainted.  With the high-density  
connectors and need for tight torque specs, it makes sense to make an  
easily-to-replace unit, saving downtime and possible damage to  
working components.

Sun's biggest problem is that they engineer the hell out of the  
systems, then get crappy components.  Case in point: We lose fan RPM  
sensors on V210 power supply fans on a semi-regular basis.  My  
educated guess is that, much like the CPU fans on the Ultra 1s, there  
is a serious supply-side issue, or a failure to take into account the  
level of heat in many crowded racks.

Or, it's just a basic design problem.  I've worked hand-in-hand with  
the folks who designed the E10K and the F15/E25K.  These are some  
brilliant people with an amazing knack for making robust and stable  
hardware.  However, I always wonder why nobody originally noticed  
that the air filters on the E10K  required a power-down to replace.   
Kind of a faux pas on a multi-million-dollar box that was supposed to  
be able to stay up for very long times.  Yeah, you had multipath I/O  
and redundancy all over the place, but you can't replace the bloody  
air filters without turning it off?  (This was fixed in the F15K, by  
the way.)

Now, if you'll excuse me, I'm going over to pat my E450 on the head.   
For a machine that turns 10 in a couple of months, it's doing very well.

-Pete



More information about the rescue mailing list