I agree with Hoff. While the EVA line was well engineered for the day, it
is really old now. I checked, and the first hardware EVA 6000 manual was
dated 2005; the 3/5000 line is at least 2 years older. My client dumped
their EVA600 around 2013. As I said, field service would performed
scheduled replacement of the batteries, and I think we still had a failure
once due to a dead battery. Depending on the projected life of the current
cluster (and there's not much information about that), it might be possible
to replace a whole rackful (or more) of servers and EVA5000 with an
RX2800i6 (one, plus possibly a second to back it up), and the SSD's Steve
loves; and I agree. cheaper options.
On Tue, Sep 29, 2020 at 3:05 PM Stephen Hoffman via Info-vax <
Post by Stephen Hoffman
... but it is incomprehensible to me that a device like this isn't able
to run in an uncached mode.
It'd have been nice to get no-cache read-only access, yes. Not a good
failure mode. Bad controller design, or maybe bad or missing doc.
I've seen various dual- and multi-redundant storage configurations fail
similarly over the years, when the earlier failures aren't noticed.
Old hardware is fine, so long as the site has spare parts, and the
ability and budget and time and skills to self-maintain.
I'd wager that that EVA will probably be replaced by a shelf of SSDs
with a massive I/O boost, if OpenVMS continues to be used here.
The failure chain here has a number of obvious links, and all of this
among the details that new director will be considering.
Pure Personal Opinion | HoffmanLabs LLC
Info-vax mailing list