[Fwd: Re: [ntp:hackers] Weekend error report]

David L. Mills mills at udel.edu
Wed Oct 6 06:20:55 PDT 2004


Bridgeport and baldwin mount pogo disks and do not build ntpd. They are 
strictly clients of pogo and use its binaries. At the moment bridgeport 
is stuck for soem reason, but its twin baldwin is okay. The lack of 
response with broadcst is also seen on the backroom wire.


Danny Mayer wrote:

> At 12:47 AM 10/6/2004, David L. Mills wrote:
> Danny,
>>> About the build. Originally, bridgeport and baldwin were to mount 
>>> pogo:/usr/local, but I see that is not the case. However, I see 
>>> evidence somebody (you?) built on bridgeport.
> It was the case when I used it. However, the disk is full on 
> bridgeport so any build
> is likely to fail. I didn't build on bridgeport but since the pogo 
> disk was mounted
> I used the pogo build to test on bridgeport. I suspect your build 
> failed because
> of the disk space problem.
> When I ran ntpd on bridgeport I didn't see any broadast packets arrive 
> even
> though rackety was supposedly acting as a broadcast server. I couldn't 
> tell
> why, or if I waited long enough. How frequently is a broadcast packet 
> supposed
> to go out? It was receiving packets from the other servers listed.
> Are there any other nodes receiving broadcast packets?
> Danny
>>> Dave
>>> At 12:37 AM 9/27/2004, Danny Mayer wrote:
>>>>> At 10:59 PM 9/26/2004, David L. Mills wrote:
>>>>>> Danny,
>>>>>> At the moment on the campus wire, rackety is broadcasting but 
>>>>>> bridgeport is not receiving.
>>>>> I just took over bridgeport to see what was happening. It put it 
>>>>> back. There
>>>>> are a number of problems:
>>>>> 1) It's out of disk space so it can't write anything to disk.
>>>>> 2) It's running an older build dated Sep 2 and not what you 
>>>>> installed elsewhere.
>>>>> 3) It's not receiving broadcast packets at all from what I can 
>>>>> tell. It does show
>>>>> that it was enabled to receive broadcast packets. How often does a 
>>>>> broadcast
>>>>> packet get sent? I only had a few minutes on which may not have been
>>>>> long enough. Are you sure rackety is sending broadcast packets?
>>>>>>  If albert is segfaulting somewhere in the address prettyprint 
>>>>>> routine and porkypine comes bum in that area, could there be a 
>>>>>> connection? The gdb bt got nowhere near the resolver library. 
>>>>>> Problem is, ntpq can't print addresses either. That may be a clue.
>>>>> We don't know what the problem is with albert except it was 
>>>>> segfaulting when
>>>>> trying to get addresses. Harlan may have something to say here 
>>>>> when he
>>>>> gets back. porkypine is, well, porkypine. Although it runs, 
>>>>> everything I tried
>>>>> to get an address back fails. This being an ancient box, I'm not 
>>>>> sure how
>>>>> hard we should try. Every idea I had failed.
>>>>>> Both Linux mort and Freebie howland eventually resolved there 
>>>>>> differences with rackety autokey, but I don't know why it took 
>>>>>> several repeats of the signature request. Poor SunOS rackety 
>>>>>> might be running out of steam.
>>>>> That could be. Is it that overloaded? I'm not sure why it would 
>>>>> require several
>>>>> tries unless the requests are timing out. Could that be what is 
>>>>> happening?
>>>>> Danny

More information about the hackers mailing list