[VOTE] Traffic Control RELEASE-1.8.0-RC3

[Dan Kirkwood]

Hello All,

I've prepared another release for v1.8.0 (RC3)

Changes since 1.7.0:
https://github.com/apache/incubator-trafficcontrol/compare/RELEASE-1.7.0...RELEASE-1.8.0-RC3

This corresponds to git:
Hash: daf585eacdcae4f57d60f14b4b6170b004058559
Tag: RELEASE-1.8.0-RC3

Which can be verified with the following:

git tag -v RELEASE-1.8.0-RC3

My code signing key is available here:
  http://keys.gnupg.net/pks/lookup?search=0x4587A8F0=vindex

Make sure you refresh from a key server to get all relevant signatures.

Note that we are not providing the RPM files this time.   The only
artifact provided is a source tar file which can be downloaded from
here:
  https://dist.apache.org/repos/dist/dev/incubator/trafficcontrol/1.8.0/RC3/

Let me know if you need the rpm files and I can make arrangements to
get them to you.

Per Apache guidelines, the .tar.gz file is signed with my pgp key and
an md5 checksum is also provided there.

With this quick turnaround since the prior RC2,  I'd like to keep the
vote open until the same date:
The vote is open until Wednesday, December 7, 2016.

Thanks!

Re: [VOTE] Traffic Control RELEASE-1.8.0-RC2

[David Neuman]

If I remember correctly, the RPMs were included as a convenience.  I am ok
with not including them, if someone wants an RPM they are easy enough to
build with the build script.

On Thu, Dec 1, 2016 at 1:40 PM, Dan Kirkwood  wrote:

> I'd also love to ditch the RPMs,but I'll abstain from voting since
> it directly impacts me immediately (less work for me!).
>
> Would anyone else like to weigh in on this?
>
> Dan
>
> On Thu, Dec 1, 2016 at 12:52 PM, Leif Hedstrom  wrote:
> >
> >> On Dec 1, 2016, at 12:46 PM, Phil Sorber  wrote:
> >>
> >> http://www.apache.org/dev/release-signing.html#basic-facts
> >>
> >> Missing checksums for the artifacts.
> >>
> >> And for the record, I am still not liking the RPM's as release
> artifacts,
> >> but I'll let the IPMC weigh in on that.
> >
> >
> > If I had a vote, now that you have the tar-ball, I’d ditch all he RPMs.
> If someone needs the RPMs, make a Makefile target such that someone can
> produce those source RPMs (shouldn’t they be .srpm) from the tar-ball.
> >
> > Cheers,
> >
> > — Leif
> >
>

Re: [VOTE] Traffic Control RELEASE-1.8.0-RC2

[Dan Kirkwood]

I'd also love to ditch the RPMs,but I'll abstain from voting since
it directly impacts me immediately (less work for me!).

Would anyone else like to weigh in on this?

Dan

On Thu, Dec 1, 2016 at 12:52 PM, Leif Hedstrom  wrote:
>
>> On Dec 1, 2016, at 12:46 PM, Phil Sorber  wrote:
>>
>> http://www.apache.org/dev/release-signing.html#basic-facts
>>
>> Missing checksums for the artifacts.
>>
>> And for the record, I am still not liking the RPM's as release artifacts,
>> but I'll let the IPMC weigh in on that.
>
>
> If I had a vote, now that you have the tar-ball, I’d ditch all he RPMs. If 
> someone needs the RPMs, make a Makefile target such that someone can produce 
> those source RPMs (shouldn’t they be .srpm) from the tar-ball.
>
> Cheers,
>
> — Leif
>

Re: [VOTE] Traffic Control RELEASE-1.8.0-RC2

[Leif Hedstrom]

> On Dec 1, 2016, at 12:46 PM, Phil Sorber  wrote:
> 
> http://www.apache.org/dev/release-signing.html#basic-facts
> 
> Missing checksums for the artifacts.
> 
> And for the record, I am still not liking the RPM's as release artifacts,
> but I'll let the IPMC weigh in on that.


If I had a vote, now that you have the tar-ball, I’d ditch all he RPMs. If 
someone needs the RPMs, make a Makefile target such that someone can produce 
those source RPMs (shouldn’t they be .srpm) from the tar-ball.

Cheers,

— Leif

Re: [VOTE] Traffic Control RELEASE-1.8.0-RC2

[Phil Sorber]

http://www.apache.org/dev/release-signing.html#basic-facts

Missing checksums for the artifacts.

And for the record, I am still not liking the RPM's as release artifacts,
but I'll let the IPMC weigh in on that.

On Thu, Dec 1, 2016 at 12:36 PM Leif Hedstrom  wrote:

>
> > On Nov 30, 2016, at 10:56 AM, Dan Kirkwood  wrote:
> >
> > Hello All,
> >
> > I've prepared a release for v1.8.0 (RC2)
> >
> > Changes since 1.7.0:
> >
> https://github.com/apache/incubator-trafficcontrol/compare/RELEASE-1.7.0...RELEASE-1.8.0-RC2
> >
> > This corresponds to git:
> > Hash: 8766dbcb38105fbc97b955b4733defe40c83db00
> > Tag: RELEASE-1.8.0-RC2
> >
> > Which can be verified with the following:
>
>
> Minor (nit-pick) detail: Why does the tar-ball unpack into a directory
> named “workspace”? I would have expected it to be
> incubator-trafficcontrol-1.8.0.4567.8766dbcb .
>
> Also, why the UUID in the release name? Is that something the incubator
> wants now?
>
> Cheers,
>
> — leif
>
>

Re: [VOTE] Traffic Control RELEASE-1.8.0-RC2

[Dan Kirkwood]

ahh..thanks,  Leif..I didn't realize it got the workspace in
there..That's a relic of our Jenkins CI build -- it clones into a
"workspace" dir.   Looks like I'll need to modify the build script to
be explicit when creating the tarball..

As for including the git hash in the file name,  we've been following
that convention in the rpm name for a while.   Do you think it should
be dropped from the tarball?   That's written to BUILD_NUMBER before
it's created..

-Dan

On Thu, Dec 1, 2016 at 12:35 PM, Leif Hedstrom  wrote:
>
>> On Nov 30, 2016, at 10:56 AM, Dan Kirkwood  wrote:
>>
>> Hello All,
>>
>> I've prepared a release for v1.8.0 (RC2)
>>
>> Changes since 1.7.0:
>> https://github.com/apache/incubator-trafficcontrol/compare/RELEASE-1.7.0...RELEASE-1.8.0-RC2
>>
>> This corresponds to git:
>> Hash: 8766dbcb38105fbc97b955b4733defe40c83db00
>> Tag: RELEASE-1.8.0-RC2
>>
>> Which can be verified with the following:
>
>
> Minor (nit-pick) detail: Why does the tar-ball unpack into a directory named 
> “workspace”? I would have expected it to be 
> incubator-trafficcontrol-1.8.0.4567.8766dbcb .
>
> Also, why the UUID in the release name? Is that something the incubator wants 
> now?
>
> Cheers,
>
> — leif
>

Re: Enhancement: Multi Delivery Services With Same Domain Name and Different Path Prefixes

[Jan van Doorn]

So for your example you would enter 2 delvieryservices with the same host_regex 
(which would be possible because you drop the unique requirement on it), 
different path prefixes and have different settings for each? 

I think I get that ?

I _think_ this would work without changing Traffic Router (it just tags on the 
path in the redirect)... 3.3 in the doc says Traffic Router will be changed as 
well, but I don't see that in the PR? 

Also, going back to my initial question - did you consider implementing this 
with PATH_REGEXP and regex_remap? 

Rgds,
JvD


> On Dec 1, 2016, at 07:35, Jifeng Yang (jifyang)  wrote:
> 
> Hi JvD,
> 
> The difference between the two is: the former doesn’t serve the content under 
> the paths other than “/vod/” and “/live/”.
> 
> For example, for the request 
> “http://traffic-server.sports.ipcdn.com/path/file”, the former doesn’t serve 
> it while the latter does serve it.
> 
> Regarding the use case, this is useful if:
> 
> Under the same domain name, the contents under some paths are set by one 
> configuration and the contents under some other paths are set by another 
> configuration.
> 
> For example,
> 
> Different “Regex remap expression” can be configured for 
> “http://traffic-server.sports.ipcdn.com/vod/” and 
> “http://traffic-server.sports.ipcdn.com/live/”.
> 
> Different traffic caches can be assigned for 
> “http://traffic-server.sports.ipcdn.com/vod/” and 
> “http://traffic-server.sports.ipcdn.com/live/”.
> 
> Some different configuration items become possible because separated delivery 
> services.
> 
> Thanks,
> Jifeng
> 
> 
> On 30/11/2016, 23:23, "Jan van Doorn"  wrote:
> 
>Hi Jifeng,
> 
>I'm still confused, bear with me please. 
> 
>The Google doc example has
> 
>maphttp://traffic-server.sports.ipcdn.com/vod/ 
> http://origin.server.com/vod/
>maphttp://traffic-server.sports.ipcdn.com/live/ 
> http://origin.server.com/live/
> 
> 
>But, isn't that the same as 
> 
>maphttp://traffic-server.sports.ipcdn.com/ 
> http://origin.server.com/
> 
>?
> 
>If you want to send the /live to RAM and the /vod to disk, they can't be 
> in the same deliveryservice table entry, since all those type if things are 
> set there? 
> 
>Can you elaborate on the use case you are trying to solve?
> 
>Rgds,
>JvD
> 
>> On Nov 30, 2016, at 04:16, Jifeng Yang (jifyang)  wrote:
>> 
>> Hi,
>> 
>> Different delivery services can be configured with different domain names 
>> now. In some cases, different delivery services with same domain name and 
>> different path prefixes are needed. These delivery services can have 
>> different configurations.
>> 
>> The problem and a solution are described in the document 
>> https://docs.google.com/document/d/19-TZ6ODla_vdiYqZajbpRiOpLvpbJxil1SvIm44zb-0/edit?usp=sharing.
>> 
>> The issue and PR for this:
>> Issue: https://issues.apache.org/jira/browse/TC-55?jql=project%20%3D%20TC
>> PR: https://github.com/apache/incubator-trafficcontrol/pull/108
>> 
>> Thanks,
>> Jifeng
>> 
> 
> 
> 

Re: Enhancement: Multi Delivery Services With Same Domain Name and Different Path Prefixes

[Jifeng Yang (jifyang)]

Hi JvD,

The difference between the two is: the former doesn’t serve the content under 
the paths other than “/vod/” and “/live/”.

For example, for the request 
“http://traffic-server.sports.ipcdn.com/path/file”, the former doesn’t serve it 
while the latter does serve it.

Regarding the use case, this is useful if:

Under the same domain name, the contents under some paths are set by one 
configuration and the contents under some other paths are set by another 
configuration.

For example,

Different “Regex remap expression” can be configured for 
“http://traffic-server.sports.ipcdn.com/vod/” and 
“http://traffic-server.sports.ipcdn.com/live/”.

Different traffic caches can be assigned for 
“http://traffic-server.sports.ipcdn.com/vod/” and 
“http://traffic-server.sports.ipcdn.com/live/”.

Some different configuration items become possible because separated delivery 
services.

Thanks,
Jifeng


On 30/11/2016, 23:23, "Jan van Doorn"  wrote:

Hi Jifeng,

I'm still confused, bear with me please. 

The Google doc example has

map http://traffic-server.sports.ipcdn.com/vod/ 
http://origin.server.com/vod/
map http://traffic-server.sports.ipcdn.com/live/ 
http://origin.server.com/live/


But, isn't that the same as 

map http://traffic-server.sports.ipcdn.com/ http://origin.server.com/

?

If you want to send the /live to RAM and the /vod to disk, they can't be in 
the same deliveryservice table entry, since all those type if things are set 
there? 

Can you elaborate on the use case you are trying to solve?

Rgds,
JvD

> On Nov 30, 2016, at 04:16, Jifeng Yang (jifyang)  
wrote:
> 
> Hi,
> 
> Different delivery services can be configured with different domain names 
now. In some cases, different delivery services with same domain name and 
different path prefixes are needed. These delivery services can have different 
configurations.
> 
> The problem and a solution are described in the document 
https://docs.google.com/document/d/19-TZ6ODla_vdiYqZajbpRiOpLvpbJxil1SvIm44zb-0/edit?usp=sharing.
> 
> The issue and PR for this:
> Issue: https://issues.apache.org/jira/browse/TC-55?jql=project%20%3D%20TC
> PR: https://github.com/apache/incubator-trafficcontrol/pull/108
> 
> Thanks,
> Jifeng
>