Guido Günther wrote:
> Hi Cole,
> On Thu, Oct 09, 2008 at 12:45:51PM -0400, Cole Robinson wrote:
>>> + def in_out_vector_limit(self, data, limit):
>>> + l = len(data)/2
>>> + end = [l, limit][l > limit]
>>> + if l > limit:
>>> + data = data[0:end] + data[l:l+end]
>> This piece here was giving me some trouble. If we
>> shrink data here, then the line below will try to
>> grab an out of bound index.
> That's broken. Fixed in the updated patch series as you suggested.
>
> [..snip..]
>> Also, the scaling problems with the disk polling can be
>> _really_ bad (though it's no fault of the code). If I have
>> polling going once a second, and single guest running with
>> six disks and a nic, the UI is completely locked up.
> Doesn't look so bad here, but I agree that this can slow down things a
> lot with plenty of machines - I made this tunable now, see below.
>
>> So I'd like to hold off on committing the actual polling
>> work until we put the gconf values in to disable this. It
>> doesn't even need to wired up to the gui for now, I can
>> do that after this its committed if you'd like (the prefs
>> dialog needs an overhaul anyways, but that can come after).
> I didn't want to introduce another gconf value if the gui needs an
> overwhole anyway so I simply used the existing
> vmlist-fiels/{network_traffic, disk_usage}. Are you o.k. with that?
>
Yep, that's fine for now. I'm probably going to make this
more fine grained in the near future, so that disabling
the column view in the manager window _doesn't_ mean skip
the polling: that will be a separate option.
My guess is, when we cut a new release, I will keep the disk
and net stats disabled as they are now, with a warning in the
prefs dialog that there may be scaling issues.
Anyways, I've applied the remaining patches. Thanks a lot
for the contribution!
- Cole
_______________________________________________
et-mgmt-tools mailing list
et-mgmt-tools@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/et-mgmt-tools
