getWorkerInfo fails during worker startup

[cmumford]

The worker was able to start and do a successful build, but cannot reconnect to the master during any restart.

Master Versions:

• Python version: 3.4.2
• Buildbot version: 0.9.10
• Twisted version: 17.5.0

Worker Versions:

• Python version: 3.5.1
• Buildbot version: 0.9.10
• Twisted version: 17.5.0

From master/twistd.log:

...
2017-09-06 13:55:29+0000 [-] buildbotNetUsageData: buildbot.net said: ok
2017-09-06 13:55:40+0000 [Broker,0,10.0.9.11] worker 'macOS-1' attaching from IPv4Address(TCP, '10.0.9.11', 54424)
2017-09-06 13:55:40+0000 [Broker,0,10.0.9.11] Failed to communicate with worker 'macOS-1'
	'ascii' codec can't decode byte 0x9f in position 3: ordinal not in range(128)
2017-09-06 13:55:40+0000 [Broker,0,10.0.9.11] Peer will receive following PB traceback:
2017-09-06 13:55:40+0000 [Broker,0,10.0.9.11] Unhandled Error
	Traceback (most recent call last):
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/twisted/internet/defer.py", line 500, in errback
	    self._startRunCallbacks(fail)
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/twisted/internet/defer.py", line 567, in _startRunCallbacks
	    self._runCallbacks()
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/twisted/internet/defer.py", line 653, in _runCallbacks
	    current.result = callback(current.result, *args, **kw)
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/twisted/internet/defer.py", line 1442, in gotResult
	    _inlineCallbacks(r, g, deferred)
	--- <exception caught here> ---
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/twisted/internet/defer.py", line 1384, in _inlineCallbacks
	    result = result.throwExceptionIntoGenerator(g)
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/twisted/python/failure.py", line 393, in throwExceptionIntoGenerator
	    return g.throw(self.type, self.value, self.tb)
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/buildbot/worker/protocols/pb.py", line 76, in _getPerspective
	    accepted = yield workers.newConnection(conn, workerName)
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/twisted/internet/defer.py", line 1384, in _inlineCallbacks
	    result = result.throwExceptionIntoGenerator(g)
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/twisted/python/failure.py", line 393, in throwExceptionIntoGenerator
	    return g.throw(self.type, self.value, self.tb)
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/buildbot/worker/manager.py", line 132, in newConnection
	    info = yield conn.remoteGetWorkerInfo()
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/twisted/internet/defer.py", line 1384, in _inlineCallbacks
	    result = result.throwExceptionIntoGenerator(g)
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/twisted/spread/pb.py", line 513, in throwExceptionIntoGenerator
	    return g.throw(RemoteError(self.type, self.value, self.traceback))
	  File "/home/chris/buildbot/sandbox/lib/python3.4/site-packages/buildbot/worker/protocols/pb.py", line 201, in remoteGetWorkerInfo
	    info = yield self.mind.callRemote('getWorkerInfo')
	twisted.spread.pb.RemoteError: 'ascii' codec can't decode byte 0x9f in position 3: ordinal not in range(128)

[seankelly]

Byte 0x9f isn't defined in ISO-8859-1 and in windows-1252 it's Ÿ. Are there any files with non-ASCII names in the info directory for the worker?

[cmumford]

I did do the first build with the stock two info files (admin and host). After the first build I edited both files (using MacVim) and restarted the worker. Once the failures started I created a second worker (using buildbot-worker create-worker worker localhost example-worker pass) and copied those two files over my edited ones, and after restart was still seeing the problems. I'll verify that no other files (like vim temp files) somehow crept into that directory after I edited it when I get back to that machine in about six hours.

[cmumford]

@seankelly yup - that was it. Temporary files. Out of curiosity, why does the implementation send over all files in the info directory? Would it be better to only send over known files?

ls -la
total 56
drwxr-xr-x  6 chris  staff    204 Sep  6 06:57 .
drwxr-xr-x  8 chris  staff    272 Sep  6 07:27 ..
-rw-r--r--  1 chris  staff  12525 Sep  5 17:21 .admin.un~
-rw-r--r--  1 chris  staff   1042 Sep  5 17:21 .host.un~
-rw-r--r--  1 chris  staff     43 Sep  6 07:06 admin
-rw-r--r--  1 chris  staff     49 Sep  6 07:06 host

[seankelly]

That code pre-dates the import from CVS in 2005. It's mentioned in the changelog though from 2003:

buildbot/bot.py (Bot.remote_getSlaveInfo): add routines to get "slave information" from $(slavedir)/info/* . These files are maintained by the slave administrator, and describe the machine/environment that is hosting the slave. Information from them is put into the "Builder" HTML page. Still need to establish a set of well-known filenames and meanings for this data: at the moment, all info/* files are sent to the master, but only 'admin' and 'host' are used on that end.