ARM buildbot not connecting

Dustin J. Mitchell dustin at v.igoro.us
Tue Oct 6 01:04:33 UTC 2015


You can see a little more detail about the communication patterns here:
  http://docs.buildbot.net/current/developer/master-slave.html

No, it all occurs over a single TCP connection from the buildslave to the
buildmaster on port 9989.

master/buildbot/buildslave/manager.py
126             yield conn.remotePrint(message="attached")
127             info = yield conn.remoteGetSlaveInfo()
128             log.msg("Got slaveinfo from '%s'" % buildslaveName)

line 126 is what results in the "message from master: attached" log on the
buildslave, so you know that is being executed.  But apparently the
`remoteGetSlaveInfo` call never completes.

So either that RPC call is never reaching the buildslave, or no response is
ever reaching the buildmaster.  You could distinguish those possibilities
by editing buildslave/base.py to add some `print` statements to
`remote_getSlaveInfo`.  At a guess, I wonder if `multiprocessing.cpu_count`
hangs on the ARM platform for some reason.

Dustin


On Mon, Oct 5, 2015 at 3:53 PM, Elliot Saba <staticfloat at gmail.com> wrote:

> To give a small update on this, it looks like some kind of failure is
> going on in the initial communication between master and slave.  Contrast
> the twistd.log file when a working slave connects:
>
> 2015-10-05 19:49:06+0000 [Broker,15,aaa.bbb.ccc.ddd] slave 'centos6.7-x64'
> attaching from IPv4Address(TCP, 'aaa.bbb.ccc.ddd', 53629)
> 2015-10-05 19:49:06+0000 [Broker,15,aaa.bbb.ccc.ddd] Got slaveinfo from
> 'centos6.7-x64'
> 2015-10-05 19:49:06+0000 [Broker,15,aaa.bbb.ccc.ddd] Starting buildslave
> keepalive timer for 'centos6.7-x64'
> 2015-10-05 19:49:06+0000 [Broker,15,aaa.bbb.ccc.ddd] bot attached
> 2015-10-05 19:49:06+0000 [Broker,15,aaa.bbb.ccc.ddd] Buildslave
> centos6.7-x64 attached to nuke_centos6.7-x64
> 2015-10-05 19:49:06+0000 [Broker,15,aaa.bbb.ccc.ddd] Buildslave
> centos6.7-x64 attached to nightly_cxx64
> 2015-10-05 19:49:06+0000 [Broker,15,aaa.bbb.ccc.ddd] Buildslave
> centos6.7-x64 attached to clean_centos6.7-x64
>
> Versus when my new ARM slave connects:
> 2015-10-05 19:49:16+0000 [Broker,16,www.xxx.yyy.zzz] slave
> 'ubuntu14.04-armv7l' attaching from IPv4Address(TCP, 'www.xxx.yyy.zzz',
> 47946)
>
> It never says "Got slaveinfo from ubuntu14.04-armv7l", which seems to me
> like there's a problem with duplex communication.  Does anyone know if I
> need to open ports other than the default buildmaster port of 9989?  Do I
> need to open ports on the buildslave side?
> -E
>
> On Fri, Oct 2, 2015 at 11:47 PM, nachaat hassis <nachaat05 at yahoo.fr>
> wrote:
>
>> The fact that the slave is an arm-board does not matter.
>> I have many arm- and armhf- slaves and they are working fine.
>>
>> Von meinem iPhone gesendet
>>
>> Am 03.10.2015 um 08:21 schrieb Elliot Saba <staticfloat at gmail.com>:
>>
>> Hello all,
>>
>> I have an ubuntu14.04 ARM machine that I'm trying to connect to my
>> buildmaster, and although both twistd.log files show the connection being
>> made, the web UI always says that the buildslave is disconnected, and no
>> jobs are ever dispatched to the buildslave.
>>
>> In slightly more detail, my buildmaster has logs of the form:
>>
>> 2015-10-03 06:16:42+0000 [Broker,12,www.xxx.yyy.zzz] slave
>> 'ubuntu14.04-armv7l' attaching from IPv4Address(TCP, 'aaa.bbb.ccc.ddd',
>> 47418)
>>
>> And my buildslave shows:
>>
>> 2015-10-03 06:16:41+0000 [-] Connecting to buildbot.e.ip.saba.us:9989
>> 2015-10-03 06:16:41+0000 [Broker,client] message from master: attached
>>
>> You can see the buildslave on the webui here
>> <http://buildbot.e.ip.saba.us:8010/builders/package_tarballarm7vl>, what
>> can I do to figure out why this is happening?  Could this have something to
>> do with the fact that the buildslave is an ARM machine?
>>
>> Thanks,
>> -Elliot
>>
>> _______________________________________________
>> users mailing list
>> users at buildbot.net
>> https://lists.buildbot.net/mailman/listinfo/users
>>
>>
>
> _______________________________________________
> users mailing list
> users at buildbot.net
> https://lists.buildbot.net/mailman/listinfo/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.buildbot.net/pipermail/users/attachments/20151005/ae9f9324/attachment.html>


More information about the users mailing list