You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@trafficcontrol.apache.org by Robert O Butts <ro...@apache.org> on 2021/08/12 16:15:07 UTC
Remove t3c (nee ORT) dispersion?
I'd like to propose we remove the t3c (nee ORT) --dispersion and
--login-dispersion flags.
These were added to deal with Traffic Ops not being able to handle the load
of all the caches in a large CDN requesting at once. Which is still true,
but now that t3c is very fast (the bottleneck now is TO requests, which
even on a very large CDN with full DeliveryServiceServers is <30s), I'm not
seeing an advantage over `sleep "$((RANDOM % (60 * 15)))" && t3c ...`
Since you can't run multiple instances, this also required a great deal of
code around "revalidate-while-sleeping" which could also be removed,
reducing dev costs and potential bugs.
Is anyone else using the ORT/t3c --dispersion or --login-dispersion flags?
Does anyone object to removing them?
Thanks,
Re: Remove t3c (nee ORT) dispersion?
Posted by Derek Gelinas <mr...@gmail.com>.
+1
On Aug 12, 2021, 5:33 PM -0400, John Rushford <jj...@gmail.com>, wrote:
> +1
>
> Sent from my iPhone
>
> > On Aug 12, 2021, at 2:46 PM, ocket 8888 <oc...@gmail.com> wrote:
> >
> > +1 that sounds much simpler
> >
> > > On Thu, Aug 12, 2021 at 10:15 AM Robert O Butts <ro...@apache.org> wrote:
> > >
> > > I'd like to propose we remove the t3c (nee ORT) --dispersion and
> > > --login-dispersion flags.
> > >
> > > These were added to deal with Traffic Ops not being able to handle the load
> > > of all the caches in a large CDN requesting at once. Which is still true,
> > > but now that t3c is very fast (the bottleneck now is TO requests, which
> > > even on a very large CDN with full DeliveryServiceServers is <30s), I'm not
> > > seeing an advantage over `sleep "$((RANDOM % (60 * 15)))" && t3c ...`
> > >
> > > Since you can't run multiple instances, this also required a great deal of
> > > code around "revalidate-while-sleeping" which could also be removed,
> > > reducing dev costs and potential bugs.
> > >
> > > Is anyone else using the ORT/t3c --dispersion or --login-dispersion flags?
> > >
> > > Does anyone object to removing them?
> > >
> > > Thanks,
> > >
Re: Remove t3c (nee ORT) dispersion?
Posted by John Rushford <jj...@gmail.com>.
+1
Sent from my iPhone
> On Aug 12, 2021, at 2:46 PM, ocket 8888 <oc...@gmail.com> wrote:
>
> +1 that sounds much simpler
>
>> On Thu, Aug 12, 2021 at 10:15 AM Robert O Butts <ro...@apache.org> wrote:
>>
>> I'd like to propose we remove the t3c (nee ORT) --dispersion and
>> --login-dispersion flags.
>>
>> These were added to deal with Traffic Ops not being able to handle the load
>> of all the caches in a large CDN requesting at once. Which is still true,
>> but now that t3c is very fast (the bottleneck now is TO requests, which
>> even on a very large CDN with full DeliveryServiceServers is <30s), I'm not
>> seeing an advantage over `sleep "$((RANDOM % (60 * 15)))" && t3c ...`
>>
>> Since you can't run multiple instances, this also required a great deal of
>> code around "revalidate-while-sleeping" which could also be removed,
>> reducing dev costs and potential bugs.
>>
>> Is anyone else using the ORT/t3c --dispersion or --login-dispersion flags?
>>
>> Does anyone object to removing them?
>>
>> Thanks,
>>
Re: Remove t3c (nee ORT) dispersion?
Posted by ocket 8888 <oc...@gmail.com>.
+1 that sounds much simpler
On Thu, Aug 12, 2021 at 10:15 AM Robert O Butts <ro...@apache.org> wrote:
> I'd like to propose we remove the t3c (nee ORT) --dispersion and
> --login-dispersion flags.
>
> These were added to deal with Traffic Ops not being able to handle the load
> of all the caches in a large CDN requesting at once. Which is still true,
> but now that t3c is very fast (the bottleneck now is TO requests, which
> even on a very large CDN with full DeliveryServiceServers is <30s), I'm not
> seeing an advantage over `sleep "$((RANDOM % (60 * 15)))" && t3c ...`
>
> Since you can't run multiple instances, this also required a great deal of
> code around "revalidate-while-sleeping" which could also be removed,
> reducing dev costs and potential bugs.
>
> Is anyone else using the ORT/t3c --dispersion or --login-dispersion flags?
>
> Does anyone object to removing them?
>
> Thanks,
>