[request-sponsor] bugid 6207437

[frank.hofm...@sun.com]

On Thu, 10 Jan 2008, Visakh M R wrote:

> fix attached.
>
>
> This message posted from opensolaris.org
>

Hi Visakh,

given the description of the bug, the submitter seems to mean something 
else by "merged" than "sum of either read or write". In fact, there's an 
explicit reference in the bug submission to the Linux iostat:

http://linuxreviews.org/man/iostat/

If I understand that right, then iostat on Linux allows observability down 
into driver implementation details - namely, to check whether the storage 
architecture managed to coalesce two or more different I/O requests (as 
seen from userland) into a single actual hardware/disk access.

Solaris doesn't have a metric for that, i.e. no kstat or anything that'd 
allow to measure this. Enabling the Linux' iostat rrqm/s and wrqm/s 
metrics, therefore, will, if I see it right, require more work than the 
diff you submitted.

There's another thing about the change that you've suggested - it changes 
the output format of iostat -x. One should be careful about such things, 
because there are a lot of scripts and utilities out there that rely on 
the existing format. Adding a column to the iostat output requires:
- ARC approval
- likely a new iostat flag to make this 'on demand', to prevent
  breaking existing utilities that parse iostat output

The first thing to research would be what the Linux metrics that the bug 
references actually mean, though :)


Best wishes,
FrankH.



--
No good can come from selling your freedom, not for all the gold in the world,
for the value of this heavenly gift far exceeds that of any fortune on earth.
--

[request-sponsor] bugid 6207437

[Visakh M R]

fix attached.
 
 
This message posted from opensolaris.org
-- next part --
bugid   : 6207437
synopsis: iostat should display merged results.
path: /solarisws/usr/src/cmd/stat/iostat


According to 'man iostat'-->

"   > The svc_t response time is not particularly significant when
> the  I/0  (r/s+w/s) rates are under 0.5 per second. Harmless
> spikes are fairly normal in such cases.
"

i just added a merged response("header io") in the beggining.
each time the pgm adds r/s and w/s and print it as i/o.

---
bash-3.00# diff -u iostat.c.orig iostat.c
--- iostat.c.orgTue Oct  2 20:15:57 2007
+++ iostat.cThu Jan 10 02:24:48 2008
@@ -801,25 +801,25 @@
if (doit) {
if (!do_conversions) {
if (do_raw == 0) {
-   fstr = " %6.1f %6.1f %6.1f %6.1f "
+   fstr = " %6.1f %6.1f %6.1f %6.1f %6.1f "
"%4.1f %4.1f %6.1f %3.0f "
"%3.0f ";
} else {
fstr = "%.1f,%.1f,%.1f,%.1f,%.1f,%.1f,"
-   "%.1f,%.0f,%.0f";
+   "%.1f,%.0f,%.0f,%.1f";
}
-   push_out(fstr, rps, wps, krps, kwps, avw, avr,
+   push_out(fstr, rps+wps, rps, wps, krps, kwps, 
avw, avr,
serv, w_pct, r_pct);
} else {
if (do_raw == 0) {
-   fstr = " %6.1f %6.1f %6.1f %6.1f "
+   fstr = " %6.1f %6.1f %6.1f %6.1f %6.1f "
"%4.1f %4.1f %6.1f %6.1f "
"%3.0f %3.0f ";
} else {
-   fstr = "%.1f,%.1f,%.1f,%.1f,%.1f,%.1f,"
+   fstr = 
"%.1f,%.1f,%.1f,%.1f,%.1f,%.1f,%.1f,"
"%.1f,%.1f,%.0f,%.0f";
}
-   push_out(fstr, rps, wps, krps, kwps, avw, avr,
+   push_out(fstr, rps+wps, rps, wps, krps, kwps, 
avw, avr,
wserv, rserv, w_pct, r_pct);
}
}
@@ -1279,7 +1279,7 @@
/* without -r option */
(void) snprintf(disk_header,
sizeof (disk_header),
-   "%-*.*sr/%cw/%c   "
+   "%-*.*si/or/%cw/%c   "
"%cr/%c   %cw/%c wait actv  "
"svc_t  w  b %s",
iodevs_nl, iodevs_nl, "device",
@@ -1288,7 +1288,7 @@
/* with -r option */
(void) snprintf(disk_header,
sizeof (disk_header),
-   "device,r/%c,w/%c,%cr/%c,%cw/%c,"
+   
"device,i/o,r/%c,w/%c,%cr/%c,%cw/%c,"
"wait,actv,svc_t,w,"
"b,%s",
ch, ch, iosz, ch, iosz, ch, header);
@@ -1296,11 +1296,11 @@
} else {
/* with -n option */
if (do_raw == 0) {
-   fstr = "r/%cw/%c   %cr/%c   "
+   fstr = "i/or/%cw/%c   
%cr/%c   "
"%cw/%c wait actv wsvc_t asvc_t  "
"w  b %sdevice";
} else {
-   fstr = "r/%c,w/%c,%cr/%c,%cw/%c,"
+   fstr = "i/o,r/%c,w/%c,%cr/%c,%cw/%c,"
"wait,actv,wsvc_t,asvc_t,"
"w,b,%sdevice";
}


difference in output:

bash-3.00# iostat -x 
 extended device statistics 
devicer/sw/s   kr/s   kw/s wait actv  svc_t  %w  %b 
cmdk0   

[request-sponsor] bugid 6207437

[Visakh M R]

hi,
i'm Visakh from ASE amritapuri, i'm working on 
bugid:6207437
synopsis:iostat should display merged requests
SCA no:OS0192

i've fixed it.
 
 
This message posted from opensolaris.org

[request-sponsor] bugid 6207437

[Visakh M R]

I am interested in bugid 6207437 
'll be submitting a fix by10th Jan 2008.

6207437 : iostat should display merged requests
Category:Subcategory utility:performance

I have initiated the process to get a SCA number.
 
 
This message posted from opensolaris.org