Your input on a proposal for sar -d change

[marie-noelle jeanson_1]

Hi,

Currently sar -d reports activity for disk activity and one of the values returned is r+w/s.
HP will add tape activity information in a later release, and proposes to display it as two separate values, r/s and w/s.
However the output of the command will show both tapes and disks activity and it would be redundant to display both r+w/s, r/s and w/s, so we would like to validate if it would be ok to provide the number of r/s and the number of w/s for both tape and disk rather than displaying r+w/s only, or r+w/s, r/s and w/s (SUN displsys r/s and w/s).

Question: would it be an issue for you if we changed the current output of sar -d (starting from that future release) to display two fields r/s and w/s instead of the current r+w/s (for both disks and tapes)?
Would that be a good enhancement?
Would that break scripts?
Any other feedback on this change?

Thanks,

Marie.

[Steven E. Protter · Accepted solution]

Any data presentation changes is likely to break scripts. Some people write scripts that depends on columns for the data.

I'm attaching my set of sar scripts, which I got from HP and improved(fixed bugs).

I'd like the lab to test with it, because its representitive and it would be a pain if it got broken.

SEP

[Pete Randall]

An issue? Not at all - at least for me anyway. It actually sounds like a good idea to me.


Pete

[A. Clay Stephenson]

As always, changes to utilities which have been in operation for decades will always break scripts. I would leave the -d output exactly as it is and use another option to display the separate r/s and w/s. Of course, the value of sar as a measurement tool for modern systems is questionable anyway. I very seldom use sar anymore but I always hate it when someone "improves" the output of a command that has remained unchanged for many years --- especially if sar -d zigged before and now it zags.

[marie-noelle jeanson_1]

Thanks all, I think the answer is clear.
Clay, just curious: what tools do you prefer to sar?

Marie.

[A. Clay Stephenson]

That's a dumb question for an HP employee to ask. You should always be trying to sell Glance and/or MeasureWare. Seriously, the metrics that MeasureWare builds into the kernel are far superior and more accurate (if less portable) than the decades old tools like sar.

[TwoProc]

Marie,

I really like the idea of separate data. But, to avoid breaking scripts ( not that it will break mine), why not just make it an optional flag - leaving the old display as is? I think it would about 3 cents :-) in additional coding costs to handle the command line switch.

Thanks for asking input before making the changes. I wish other unnamed vendors to be as considerate!

[Stephen Keane]

Don't forget the cheapskates/those with small budgets. We can't all justify the moolah for Glance etc. Plus som sites use "portable" scripts that run on a variety of Unices.

[marie-noelle jeanson_1]

Thanks all!

[Bill Hassell]

What's wrong with sar -t to show tape I/O. I can't imagine mixing disk and tape in the same metric, whether it is a combination with one number or separate fields. Tapes are different and it would be really useful to know what tape I/O is going on, especially with backups. It would go a long way to demonstrating data starvation for new super-tapes.