On Sat, 14 Jul 2007, Stephen Jones wrote:

>
> On Jul 14, 2007, at 9:13 AM, Anders Hogrelius wrote:
>> I loaded the latest version of the SRM firmware (6.6.10) for the DS20L on 
>> mine and it works perfectly. Tru64 even identifies the onboard temp and 
>> voltage sensors correctly which means I don't have to disable the systems 
>> health daemon anymore. With the CS20 firmware it complains about a faulty 
>> PSU and temp sensor.
>
> Wow, that is good news and quite surprising that a fix like that would be 
> incorporated this late in the game.
> Do you know what other fixes were made between 5.8-81 and 6.6.10?  I suppose 
> with a little work a consolidated
> list could be made.  I know that you can get enough info to at least to 6.0, 
> so it shouldn't take much work.
> The question is if its the newer versions actually help with making the 
> platform more stable and probably not.
>
That is if you can get info about the CS20 version of the SRM firmware. 
With the API website gone I'm not sure where to get the documentation for 
the old firmware. For a while I ran both versions in parallel with one of 
my two boxes updated. It seemed more stable but that may be an effect of 
Tru64 identifying the box as a DS20L which is supported rather than as a 
API CS20...

>> A bit more research revealed that Compaq actually made DS20L's with both 
>> the Adaptec SCSI controller as well as the Symbios Logic controller. Both 
>> types of riser cards exist for the DS20L but with different part ID's and 
>> the Compaq/DEC firmware for the DS20L identifies the Symbios Logic 
>> controller correctly. -Too bad there aren't any drivers in Tru64 for it.
>
> I thought that there was a way but installing to an IDE drive first, at least 
> this procedure seems to outline that:
> http://howto.unixdev.net/Tru64-CS20-HOWTO.html
>
Yep, thats where I found the info about how to install Tru64 on my boxes. 
I've emailed the person who wrote the page and asked him to update it 
since this info changes things a bit to the better.

>> The DS20L seems to be more than just inspired by the API CS20...
>
> I think that the understanding is that API basically built the CS20 and then 
> sold or licensed the design back
> to Compaq.  They're great machines as long as you keep them cool and their 
> fans clear.  SDF has about 24 of
> these, so we're on the platform for at least the next 5 years and it would be 
> great to be able to refurb some
> of or old PSes that have failed.  One interesting thing is that with NEB 2.1 
> we've seen the longest duration
> of uptime under an MP kernel with a significant 24/7 user load.  3.x doesn't 
> get anywhere near as close, which
> is a shame.  And its impossible to debug because its a situation where the 
> one processor hangs and if you can
> get to the debugger, you can't get any info on that hung processor.

That's good news for me as I had terrible problems with NEB 3.1 in a MP 
configuration on my DP264. It might be worth trying the older release on 
it instead. Even with a light load it would hang or crash. The DS20 and 
DP264 are similar enough to the CS20 to warrant a try...

Cheers,
Anders