[OpenAFS] sysid

[Derrick Brashear]

On Thu, Jan 13, 2011 at 2:12 PM, Jaap Winius <jwinius@umrk.nl> wrote:
> Quoting Derrick Brashear <shadow@gmail.com>:
>
>>> ... when does the system automatically recreate the sysid file, and whe=
n
>>> does it not do this? What are the requirements and can they be checked?
>>
>> if it's missing it should be recreated when the server registers with
>> the VLserver.
>>
>> what does the FileLog tell you?
>
> Oh, um... that sounds logical. :-) =A0Let's see:
>
> =A0 VL_RegisterAddrs rpc failed; The IP address exists on a different
> =A0 =A0 =A0server; repair it
> =A0 VL_RegisterAddrs rpc failed; See VLLog for details
>
> Those two lines repeated 37 times. All the VLLog said was:
>
> =A0 Using 82.95.170.57 as my primary address
> =A0 Starting AFS vlserver 4 (/usr/lib/openafs/vlserver)
>
> That's my fault. While troubleshooting a problem, I decided to add an
> internal address, 192.168.24.10, to the server's NetInfo file, in additio=
n
> to 82.95.170.57. The server's CellServDB only contains 82.95.170.57 (the
> same host also runs a ptserver and a vlserver).
>
> After removing the internal address from the NetInfo file, the errors wer=
e
> no longer present in the FileLog after the file server was restarted and =
a
> new sysid was created automatically.
>
>> f912d849d3201b004ae7dbacf2d4848836fa986f was needed such that an
>> immediate failure would be retried, but
>> 1) you should have it
>> and
>> 2) that only applied if you restarted everything (including vlserver)
>> in a single server configuration, not just the fileserver.
>
> What do you mean by "f912d849d3201b004ae7dbacf2d4848836fa986f"?

this is the problem with us ditching delta names in favor of git
sha1s. they don't roll off the tongue the same .

STABLE14-viced-vlregister-even-if-delayed-2009XXXX would have been more obv=
ious.

In any case, if you git clone the openafs git tree, that sha1
represents a commit in the 1.4 branch which fixed the issue I was
speaking of.



--=20
Derrick