Re: [Xen-devel] [PATCH] xentop: Dynamically expand some columns

["Charles Arnold" <carnold@xxxxxxxx>]

>>> On 10/3/2014 at 03:00 AM, Ian Campbell <Ian.Campbell@xxxxxxxxxx> wrote: 
> On Thu, 2014-10-02 at 13:05 -0400, Konrad Rzeszutek Wilk wrote:
>> On Thu, Oct 02, 2014 at 10:25:41AM -0600, Charles Arnold wrote:
>> > >>> On 10/2/2014 at 10:10 AM, Andrew Cooper <andrew.cooper3@xxxxxxxxxx> 
>> > >>> wrote: 
>> > > On 02/10/14 16:58, Charles Arnold wrote:
>> > >> Allow certain xentop columns to automatically expand as the amount
>> > >> of data reported gets larger.  The columns allowed to expand are:
>> > >>
>> > >> NETTX(k), NETRX(k), VBD_RD, VBD_WR, VBD_RSECT, VBD_WSECT
>> > >>
>> > >> Author: Markus Hauschild <Markus.Hauschild@xxxxxxxxxxxxxxxxxxxx>
>> > >> Signed-off-by: Charles Arnold <carnold@xxxxxxxx>
>> > > 
>> > > In principle, very nice.  (I have wanted to see about doing this for a
>> > > while now, but very far down the todo list)
>> > > 
>> > > How about the NAME field? 9 characters isn't enough for some people.
>> > 
>> > Sure, and any others that might need it.  But what to do about the '-f' 
> flag
>> > which says show me the full VM name but the flag doesn't adjust the entire
>> > column.  By automatically showing the full name and adjusting the column
>> > appropriately it makes this flag pointless (which I'm ok with). 
>> 
>> Perhaps this functionality should be under that option?
>> 
>> I am a bit hesistant about this as there are some users of xentop that
>> use it for their monitoring. What I can't remember is if they use the
>> batched mode or not - and if they scan for specific strings (and length).
>> 
>> This would (I think?) throw a wrench in that?
> 
> IMHO people who are parsing the output of tools such as this ought to be
> prepared to deal with occasional changes in the precise content of the
> output across Xen releases.
> 
> If they aren't happy with that then they should file wishlist bugs (or
> better: patches) asking for an extensible/machine-readable format to be
> output upon request, or for libxenstat to be exposed and made API
> stable, or some other functionality which solves their need.
> 
> Otherwise we find ourselves in a position where useful patches like
> Charles' and similar patches which add genuinely useful output for human
> readers can never be accepted.

So do we get a thumbs up on this?  If so can someone take the patch (I prefer 
v3 which aligns the name if -f is used, but v2 would satisfy the customer).

- Charles



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel