Moving darkserver to a staging instance
 

Hi all,

I am ready to move darkserver to a staging instance. The test instance
is working smoothly.
RFR for the same can be found at [1].

The service does not require any other Fedora service to be up. It can
use normal apache level caching.


[1] https://fedorahosted.org/fedora-infrastructure/ticket/3054


Kushal
--
http://fedoraproject.org
http://kushaldas.in
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

Moving darkserver to a staging instance
 

On Wed, 14 Mar 2012 00:34:29 +0530
Kushal Das <kushaldas@gmail.com> wrote:

> Hi all,
>
> I am ready to move darkserver to a staging instance. The test instance
> is working smoothly.
> RFR for the same can be found at [1].
>
> The service does not require any other Fedora service to be up. It can
> use normal apache level caching.
>
> [1] https://fedorahosted.org/fedora-infrastructure/ticket/3054

So, from my discussions with you on IRC, I think:

We setup a single instance for this.
We put it behind our proxy servers to allow for some caching at a local
area level.

Do we need any SOP's for this service? Is there any administration that
we might need to run on it day to day or in case of power outage, etc?

What kind of db does this instance use? Mysql? Postgres?

What kind of communication needs to happen between this and the koji
plugin? We may run into some issues here as we have a seperate builder
network that koji is on... so we will need to figure out how to get
them to communicate.

Thats all I can think of... I can make you a staging instance soon...

kevin


_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

Moving darkserver to a staging instance
 

On Wed, Mar 14, 2012 at 4:12 AM, Kevin Fenzi <kevin@scrye.com> wrote:
>
> So, from my discussions with you on IRC, I think:
>
> We setup a single instance for this.
> We put it behind our proxy servers to allow for some caching at a local
> area level.
I have a few questions on load balancing. Say if we have only one
instance and it goes down for some reason, it will not get the records
from koji during downtime.
So if we have a load balancer and more than 1 instance (say 2) , will
the load balancer push all requests to the live instance ?
If yes, then having a load balancer is a good idea.


> Do we need any SOP's for this service? Is there any administration that
> we might need to run on it day to day or in case of power outage, etc?
Only thing I know is the regular database backup.

> What kind of db does this instance use? Mysql? Postgres?
MySQL.

> What kind of communication needs to happen between this and the koji
> plugin? We may run into some issues here as we have a seperate builder
> network that koji is on... so we will need to figure out how to get
> them to communicate.
Currently the koji-plugin connects to the server using XMLRPC. They
can communicate if they can reach each other. Not sure how the current
infra works.

Kushal
--
http://fedoraproject.org
http://kushaldas.in
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

Moving darkserver to a staging instance
 

On Wed, 14 Mar 2012 11:45:40 +0530
Kushal Das <kushaldas@gmail.com> wrote:

> On Wed, Mar 14, 2012 at 4:12 AM, Kevin Fenzi <kevin@scrye.com> wrote:
> >
> > So, from my discussions with you on IRC, I think:
> >
> > We setup a single instance for this.
> > We put it behind our proxy servers to allow for some caching at a
> > local area level.
> I have a few questions on load balancing. Say if we have only one
> instance and it goes down for some reason, it will not get the records
> from koji during downtime.

yeah. After talking on IRC with you it sounds like we are going to
change things so the koji plugin is the thing that directly talks to
the db, so it shouldn't result in any loss if the frontend is down.

> So if we have a load balancer and more than 1 instance (say 2) , will
> the load balancer push all requests to the live instance ?
> If yes, then having a load balancer is a good idea.

It would.

Would this be useful for end users? Or would it be ok to have limited
downtime and they would just retry again later?

> > Do we need any SOP's for this service? Is there any administration
> > that we might need to run on it day to day or in case of power
> > outage, etc?
> Only thing I know is the regular database backup.

We do this on all our databases. ;)

They cross backup to each other, and also tape backups of all database
machines are done.

> > What kind of db does this instance use? Mysql? Postgres?
> MySQL.

ok.

> > What kind of communication needs to happen between this and the koji
> > plugin? We may run into some issues here as we have a seperate
> > builder network that koji is on... so we will need to figure out
> > how to get them to communicate.
> Currently the koji-plugin connects to the server using XMLRPC. They
> can communicate if they can reach each other. Not sure how the current
> infra works.

ok. Sounds like this won't be a problem.

kevin
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

Moving darkserver to a staging instance
 

On Thu, Mar 15, 2012 at 12:16 AM, Kevin Fenzi <kevin@scrye.com> wrote:
> On Wed, 14 Mar 2012 11:45:40 +0530
> Kushal Das <kushaldas@gmail.com> wrote:
>
>> On Wed, Mar 14, 2012 at 4:12 AM, Kevin Fenzi <kevin@scrye.com> wrote:
>> >
>> > So, from my discussions with you on IRC, I think:
>> >
>> > We setup a single instance for this.
>> > We put it behind our proxy servers to allow for some caching at a
>> > local area level.
>> I have a few questions on load balancing. Say if we have only one
>> instance and it goes down for some reason, it will not get the records
>> from koji during downtime.
>
> yeah. After talking on IRC with you it sounds like we are going to
> change things so the koji plugin is the thing that directly talks to
> the db, so it shouldn't result in any loss if the frontend is down.
>
>> So if we have a load balancer and more than 1 instance (say 2) , will
>> the load balancer push all requests to the live instance ?
>> If yes, then having a load balancer is a good idea.
>
> It would.
>
> Would this be useful for end users? Or would it be ok to have limited
> downtime and they would just retry again later?
For starting we can have only one instance. If we get enough requests
saying the users/tool can not have downtime then we might move to load
balancing.

Kushal

--
http://fedoraproject.org
http://kushaldas.in
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure