Thread: Switching to c-ares for pgbouncer?

Hi,

I was reading this:

https://redmine.postgresql.org/issues/1444

Any objections for switching to c-ares for pgbouncer? From packaging point of
view, I confirmed that c-ares is available for all supported distros, but I'm
not sure about the implications.

Regards,

--

Devrim GÜNDÜZ
Principal Systems Engineer @ EnterpriseDB: http://www.enterprisedb.com
PostgreSQL Danışmanı/Consultant, Red Hat Certified Engineer
Twitter: @DevrimGunduz , @DevrimGunduzTR

> On Jul 12, 2016, at 8:09 AM, Devrim Gündüz <devrim@gunduz.org> wrote:
>
>
> Hi,
>
> I was reading this:
>
> https://redmine.postgresql.org/issues/1444
>
> Any objections for switching to c-ares for pgbouncer? From packaging point of
> view, I confirmed that c-ares is available for all supported distros, but I'm
> not sure about the implications.

If the chart is to be believed, the only issue is:

ipv6+CNAME buggy in <=1.10

Do you know which versions are standard in the supported distros these days?

Hi Jeff,

On Tue, 2016-07-12 at 08:14 -0700, Jeff Frost wrote:
> If the chart is to be believed, the only issue is:
>
> ipv6+CNAME buggy in <=1.10
>
> Do you know which versions are standard in the supported distros these days?

RHEL 5: 1.6.0
RHEL 6, RHEL 7, Fedora 22, Fedora 23: 1.10.0
Fedora 24, rawhide: 1.11.0

Quick note: Even though CNAME+IPv6 is buggy, that feature even does not exist
in libevent (at least this is what I understand from the chart ("ipv4 only")),
so we are not introducing new bugs here.

Regards,

--

Devrim GÜNDÜZ
Principal Systems Engineer @ EnterpriseDB: http://www.enterprisedb.com
PostgreSQL Danışmanı/Consultant, Red Hat Certified Engineer
Twitter: @DevrimGunduz , @DevrimGunduzTR

> On Jul 12, 2016, at 8:31 AM, Devrim Gündüz <devrim@gunduz.org> wrote:
>
>
> Hi Jeff,
>
> On Tue, 2016-07-12 at 08:14 -0700, Jeff Frost wrote:
>> If the chart is to be believed, the only issue is:
>>
>> ipv6+CNAME buggy in <=1.10
>>
>> Do you know which versions are standard in the supported distros these days?
>
> RHEL 5: 1.6.0
> RHEL 6, RHEL 7, Fedora 22, Fedora 23: 1.10.0
> Fedora 24, rawhide: 1.11.0
>
> Quick note: Even though CNAME+IPv6 is buggy, that feature even does not exist
> in libevent (at least this is what I understand from the chart ("ipv4 only")),
> so we are not introducing new bugs here.

Sounds like at least nobody could be using ipv6 with cnames on the current setup, so I agree, it seems fine.