I've received some off-line comments from Chris regarding the iostat(1M)
change and based on this, I'd like to withdraw the iostat -n change
in behavior from this case. I'll update the materials shortly.
Should the timer be reset on this case to give people time to review
this change, and if so, what would be appropriate? Not that I'm
expecting it to be controversial, the form of name <driver><instance>
happens to be identical to the ucb form of name but is not
actually related to ucblinks or the /dev ucb names in any way.
-jg
--------------------
I think the '-n' change to iostat(1M) is the highest risk thing
you are proposing, and I am not in favor of this change as
specified.
In addition to the translation problem, there are truncation
issues to resolve. I think if you look, you will find some CRs
against current 'iostat -n' causing incorrect/truncated output.
Things look good with
# iostat -n
tty c1t0d0 c1t1d0 c0t0d0 cpu
tin tout kps tps serv kps tps serv kps tps serv us sy wt id
0 0 17 1 4 0 0 0 0 0 0 0 0 0 100
but with fabric/mpxio names current -n behavior truncates/breaks.
# iostat c3t216000C0FF8047DDd0 c3t216000C0FF8047DDd10 c3t216000C0FF8047DDd11
tty sd16 ssd178 ssd182 ssd216 cpu
tin tout kps tps serv kps tps serv kps tps serv kps tps serv us
sy wt id
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 0 99
# iostat -n c3t216000C0FF8047DDd0 c3t216000C0FF8047DDd10
c3t216000C0FF8047DDd11
tty c0t1d0 c3t216000C0FF c3t216000C0FF c3t216000C0FF cpu
tin tout kps tps serv kps tps serv kps tps serv kps tps serv us
sy wt id
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 0 99
I would prefer to see a translation mechanism delivered prior
to an iostat change. With a translation mechanism in place, you
could then revisit the -n proposal in a more limited way - one that
does not cause truncation: maybe -X implies -n, or something like that
(for line-per-device type output).
-Chris
