[nycbug-talk] Mac OSX Troubleshooting tools

[Bob Ippolito]

On Dec 18, 2004, at 8:37 PM, Pete Wright wrote:

> On Sat, Dec 18, 2004 at 01:53:02PM -0500, Bob Ippolito wrote:
>>
>> On Dec 18, 2004, at 1:43 PM, G. Rosamond wrote:
>>
>>> On Dec 18, 2004, at 1:40 PM, Charles Sprickman wrote:
>>>
>>>> On Sat, 18 Dec 2004, Isaac Levy wrote:
>>>>
>>>>> ikebook:/Users/ike ike$ iostat
>>>>> iostat: sysctl(kern.tty_nin) failed: No such file or directory
>>>>> iostat: disabling TTY statistics
>>>>>         disk0       cpu
>>>>> KB/t tps  MB/s  us sy id
>>>>> 9.76   1  0.01  13  6 81
>>>>>
>>>>> --
>>>>> Could be me- will investigate eventually...
>>>>
>>>> Not just you:
>>>>
>>>> [oof:~/Desktop] spork$ iostat
>>>> iostat: sysctl(kern.tty_nin) failed: No such file or directory
>>>> iostat: disabling TTY statistics
>>>>          disk0           disk1       cpu
>>>>  KB/t tps  MB/s   KB/t tps  MB/s  us sy id
>>>> 26.10   1  0.02   4.67   0  0.00   8  3 89
>>>> [oof:~/Desktop] spork$ uname -a
>>>> Darwin oof.local 7.6.0 Darwin Kernel Version 7.6.0: Sun Oct 10
>>>> 12:05:27 PDT 2004; root:xnu/xnu-517.9.4.obj~1/RELEASE_PPC  Power
>>>> Macintosh powerpc
>>>>
>>>> I'll have to give it a try in Tiger next time I boot to it.
>>>
>>> [gman at GMans-Computer gman]$ iostat
>>> iostat: sysctl(kern.tty_nin) failed: No such file or directory
>>> iostat: disabling TTY statistics
>>>          disk0       cpu
>>>  KB/t tps  MB/s  us sy id
>>> 15.17   8  0.11  16  6 78
>>> [gman at GMans-Computer gman]$ uname -a
>>> Darwin GMans-Computer.local 7.7.0 Darwin Kernel Version 7.7.0: Sun 
>>> Nov
>>> 7 16:06:51 PST 2004; root:xnu/xnu-517.9.5.obj~1/RELEASE_PPC  Power
>>> Macintosh powerpc
>>> [gman at GMans-Computer gman]$
>>
>> The man page says that iostat gives you some metric of system
>> statistics averaged over your uptime.  It doesn't really say much 
>> about
>> the performance of your machine, it's really more of a "how much stuff
>> did you compile today?" kind of statistic :)
>>
>
> errrr...sorta.  if you iostat over time you can gain a metric of
> transactions per sec. and kb. per trans. etc...  From those metrics 
> it's
> quite easy to deduce where potential bottlenecks may be etc.

I don't really agree.  From a little tinkering, I don't see how you 
could use iostat to find bottlenecks unless you're running at max load 
and looking at incremental statistics (but even then... I wouldn't 
really trust it).  For example, looking at the I/O while I'm playing 
iTunes isn't going to tell me where any bottleneck is, it's just going 
to tell me that my disk is getting hit once or twice for about 70k 
every two or three seconds when playing an mp3.  I have no idea if 
that's coming from cache (kernel, disk, raid, or otherwise), or even if 
it's iTunes making those transactions (probably a lot of them are 
actually from Mail, not iTunes).  It would say more or less the same 
thing on a powerbook with a 4800 RPM harddrive as it would on a dual G5 
with an Xserve RAID SAN full of 15krpm harddrives.

Then again, maybe I'm missing something, but it doesn't seem like it's 
capable of really answering the original question about the max disk 
throughput and and the performance of the RAID cache given various data 
sets.  iostat will tell you the average disk throughput of the 
applications you're running over some time interval, which is of course 
going to be somewhere between zero and the max throughput, but probably 
closer to zero :)

>> If you pass it a wait argument, *and* you are putting some serious I/O
>> load on your machine, then you might see something useful.. but it 
>> sure
>> seems like this is the wrong tool for this job.
>
> have you used iostat?  While I don't think it's by any means a "one
> stop" utility, but I don't think one would want to use one
> utility/metric for this type work.

No, I haven't ever tried to make it do anything useful.  I use other 
tools, especially Shark, but that's primarily useful for software 
developers not system architects.

-bob