Jerry Gilliam wrote:
>
> 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.
I don't think you need to do anything -- the case will probably be
approved today. I'll +1 as specified now.
- Garrett
>
>
> -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
>
