Replies inline. I've cut out some of the extraneous content to reduce
clutter

On Fri, May 24, 2019 at 3:51 AM Andrea Aime <[email protected]>
wrote:

> On Fri, May 24, 2019 at 1:35 AM Torben Barsballe via Geoserver-devel <
> [email protected]> wrote:
>
>>
>>    - Configuration can not be assumed to be persistent between builds.
>>       This will likely require the release jobs to be reconfigured somewhat.
>>
>> Ouch yes, that will definitely require the release scripts to be
> modified, as they are designed to work against a stable checkout. One
> possible approach is to make them create remote branches somewhere to
> persist the state of the work.
>
>
That seems like a reasonable solution (and honestly, is probably better
practice than having state stored on the CI server anyway).


>
>>    - geotools-<ver> and geotools-<ver>-docs builds are building
>>    successfully, but failing when they try to copy javadocs/docs
>>    (respectively) to geotools.org. We need to aquire the key to
>>    geotools.org and add it to jenkins before we can get these jobs
>>    working (or disable doc uploads until we have the key).
>>
>> Who owns that key?
>

That is one key that I know Planet Federal doesn't own. I think we had to
get in touch with OSGeo to get the key last time, but I'm not certain -
thats part of the reason why I am bringing it up. I'll dig into old list
emails and see if I can track a bit more info down.

>
>>    - Cite and Online jobs aren't going to work until PostGIS is
>>    configured on the worker nodes
>>
>> Makes sense. Is the worker node setup/configuration accessible so that
> one can help installing the packages?
> How are these managed, docker images?
>
> They are AWS AMIs with a script that runs to install the packages. The
worker nodes are one of the things that I didn't set up, and they aren't
yet accessible (I can't even get into them right now). I might have a bit
more information on this on Tuesday - the person who set the nodes up is on
holiday until then.

>
>>    - Release jobs require github and sourceforge keys, and may need some
>>    other modifications
>>    - I haven't looked at the geofence/geogig/geoscript jobs, and don't
>>    know what is needed to get them working.
>>
>> The geofence job seems to be just failing to deploy artifacts with a 401,
> checked a nightly, it failed on a corrupt geotools jar. So, nothing special
> it seems, I've
>
>>
>> The following points will require some coordination between various
>> project administrators:
>>
>>    - If you have access to the geotools.org ssh key, please contact me
>>    so we can add it to Jenkins.
>>    - If you are an administrator for both the GeoTools and GeoServer
>>    projects in GitHub, try and create a separate CI account with push
>>    permissions to those projects and generate an ssh key for it, so that the
>>    release jobs will be able to push artifacts to GitHub
>>    - If you are an administrator for both the GeoTools and GeoServer
>>    projects in SourceForge, try and create a separate CI account with push
>>    permissions to those projects and generate an ssh key for it, so that the
>>    release jobs will be able to push artifacts to SourceForge
>>
>>
> I can create these account.
> Are they going to be still managed by a single physical person (the one
> that creates the accounts and the keys), or are there better approaches?
> Searching the internet I've found nothing, suggestions welcomed.
>

Thanks.
I've got no good ideas either, and would also welcome suggestions.


>
> There are a couple of outstanding issues not discussed in the mail:
>
>    - https is not setup yet, or not working (
>    https://build.geoserver.org/geoserver does not response,
>    https://build.geoserver.org/ neither) and all links in the release are
>    pointing to it. Setting up SSL should be quick (my devops colleagues say,
>    eventually using a letsencrypt key), but if there are troubles, I can
>    switch at least the download pages to use http instead?
>
> Right, forgot about that one. It just hasn't been set up yet, still a todo.


>
>    - All builds are failing while sending the final mail due to lack of a
>    SMTP server. Thinking out loud, if needs be, I could create a gmail account
>    for the purpose, and then I guess we could make Jenkins use the gmail smtp
>    servers?
>
> While the builds throw an error, this isn't resulting in a failing status.
Most of the GeoServer jobs show success even thought they throw the SMTP
error.
I presume we had a SMTP server set up somewhere for the old server and can
probably use a similar solution here, but thats just another thing that no
one's has time to look into yet.

Cheers,

Torben
_______________________________________________
Geoserver-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/geoserver-devel

Reply via email to