Hi Andrew,

thanks for your information and advice.

I was not aware of the 'fs newcell' with 0.0.0.0
I try it without an IP, but that gives me an error-message.

/Ralf


Am 04.05.2010 um 20:57 schrieb Andrew Deason:

> On Tue, 4 May 2010 13:01:48 -0500
> Andrew Deason <[email protected]> wrote:
> 
>>> Questions:
>>>  * How can we migrate those clients to use the AFSDB-records without
>>>  rebooting (will the kernel-resisdent list updated if the afsd
>>>  started with -afsdb)?
>>>  * Or is the reboot needed to get rid of the kernel-list?
> [...] 
>> Anyway, AFSDB records will not override static information set via
>> CellServDB or 'fs newcell' (unless the CellServDB specifies no hosts).
>> So, as far as I know, you need to reboot/restart the clients.
> 
> Actually, you can use 'fs newcell' to do this, though we don't document
> that you can (and it looks like it's accidental, to me). If you give
> 0.0.0.0 as the only IP address(es) for a cell, it will effectively clear
> the list. (or if you give it a hostname that doesn't exist, it will also
> do that, but IMHO that's a bug)
> 
> So, if you do this:
> 
> fs newcell cell.name.foo 0.0.0.0
> 
> That client should use DNS lookups for dbserver information from then
> on, assuming it was started with -afsdb. If it was not started with
> -afsdb, that cell will become inaccessible until the client is
> restarted, or you provide a dbserver list with 'fs newcell' again.
> 
> This seems a bit odd, though useful. Should we document this? Or should
> we prevent this behavior, and perhaps create a new command / option to
> clear the host list for a cell?
> 
> -- 
> Andrew Deason
> [email protected]
> 
> _______________________________________________
> OpenAFS-info mailing list
> [email protected]
> https://lists.openafs.org/mailman/listinfo/openafs-info

Attachment: smime.p7s
Description: S/MIME cryptographic signature

Reply via email to