[users at bb.net] Using buildbot 0.9rc3 in production

[Elliot Saba]

Aha, it turns out I just had to finish some builds, now the console view is
working fine.  As a side note, it would be nice if the waterfall view was
sorted similarly to the console view, as it stands, the columns of the
waterfall view are sorted rather peculiarly.
-E

On Sat, Oct 1, 2016 at 1:17 AM, Elliot Saba <staticfloat at gmail.com> wrote:

> Thanks Pierre!
>
> I do not believe a firewall can be responsible for this, the two machines
> are sitting on the same local network, and I have a buildbot 0.8 instance
> with buildworkers on the same network that function just fine.  That being
> said, I destroyed the two machines and rebuilt them from the ground up on
> 0.9.0-rc4, and things seem to be working just fine, so at this point I'm
> going to blame cosmic rays and wait to see if the issues pop up again.
>
> I have a few more questions about 0.9:
>
> 1) Is it possible to arrange the rows of the table views such as the
> "Schedulers" view or the "Workers" view?  They do not seem to be sorted in
> any particular way, and it would be useful to either sort them
> alphabetically, or even better allow sorting along arbitrary columns.
>
> 2) It appears that all schedulers are listed twice.  Is this intentional,
> or am I misreading something?  See, for example, this page
> <https://buildtest.e.ip.saba.us/#/schedulers>.
>
> 3) How can I force multiple builds at once? I would expect that I can
> enter the "Schedulers" page, select one of my force schedulers and initiate
> a build across multiple buildworkers as I could in 0.8 (with the checkbox
> interface) but that doesn't seem to be the case anymore.  As it stands, I
> have to navigate to each builder page in turn and force the build there.
> This isn't very fun when you have, for instance, packaging builds you want
> to kick off for multiple OS's at once.
>
> 4) Enabling the "console" view, then performing a "reload" will leave the
> build master in an invalid state, causing the web interface to error out.
> You have to stop the buildmaster and then start it again after enabling a
> new plugin, it seems.  Should there be a bug about this?
>
> 5) The console view <https://buildtest.e.ip.saba.us/#/console> doesn't do
> anything for me; it just sits there. The log doesn't show anything
> happening, is this a bug perhaps?
>
> Thanks!
> -E
>
>
> On Wed, Sep 28, 2016 at 5:59 AM, Pierre Tardy <tardyp at gmail.com> wrote:
>
>>
>>
>> Le ven. 23 sept. 2016 à 22:45, Elliot Saba <staticfloat at gmail.com> a
>> écrit :
>>
>>> Hello Pierre, thank you for the very quick response!
>>>
>>
>> Hi Eliot,
>> thats why I try to answer quickly. If I dont, then I forgot to answer
>>
>>
>>>
>>> Using `ps -ef | grep python`, I can see that my buildbot-worker process
>>> has been running continually for ~3 days now, so I don't think the process
>>> is getting killed.  Furthermore, the buildbot-worker twistd.log shows that
>>> things are (fairly) normal, except that we seem to be hitting timeouts.  Here
>>> is a section of the buildbot-worker's log
>>> <https://gist.github.com/staticfloat/bb9cefe66704b66d69a44ba027a18409>
>>> that I believe shows an entire "build cycle".  The connection failures
>>> repeat many more times before reconnecting to the master properly, and the
>>> whole cycle repeats.
>>>
>>
>> Besides the worker process being killed, another explanation is that you
>> have some firewall that does not like long lasting connection on unknown
>> ports. Could that be the case for you?
>>
>> If the worker is doing nothing, do you also see the disconnection every
>> 16min?
>>
>>>
>>>
>>> Another question: when I navigate into detail pages, (for instance, when
>>> navigating into the build_ubuntu64 builder page
>>> <https://buildtest.e.ip.saba.us/#/builders/15>) then trying to navigate
>>> back, I click on the "Builders" breadcrumb at the top of the page, which
>>> takes me back to the "Builders" page, but doesn't get rid of the
>>> "build_ubuntu64" breadcrumb at the top.  Is that expected?
>>>
>>
>> Yes, this is a known issue related to angularjs way of doing the scope
>> lifecycle. You need to be careful, and verify that the page is still there
>> when you recieve the answer to a REST api.
>> In other term, my answer for now is: stop clicking to fast!
>> The real answer is that this is something that has to be fixed, but
>> probably not for 0.9.0
>>
>> http://trac.buildbot.net/ticket/3617
>>
>> I will be releasing rc4 tonight with the UI width fix that you requested.
>>
>> Regards
>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.buildbot.net/pipermail/users/attachments/20161003/46388eea/attachment.html>