[OpenAFS] Re: OpenAFS 1.3.62 release available

Michael Niksch nik@zurich.ibm.com
Fri, 26 Mar 2004 01:23:10 +0100


Thanks a lot for your incredibly fast response!

Jeffrey Altman wrote:
> Michael Niksch wrote:
> 
>> > For Windows, 1.3.62 represents the currently supported client. We
>> > recommend upgrading from previous releases if you are having problems,
>> > several issues from 1.3.61 have been addressed (including integrated
>> > logon issues, installer problems, and a crash on afsd service
>> > shutdown).
>>
>> Unfortunately, I am still having a couple of problems.
>>
>> 1. Stopping the service still causes a crash of afsd_service.exe.
> 
> yes.  this is a known problem. 
> we are working on it (unsuccessfully).  The problem does not occur for 
> all users
> and I cannot re-produce it.  The problem goes away if you re-compile 
> with VC 6
> instead of VS.NET 2003.  However, there are reasons why you don't want 
> to do this.
> 
>> 2. Adding a seconds drive mapping in the Drive Letters tab does not 
>> refresh the list.
> 
> Please submit a bug to openafs-bugs@openafs.org

I'll do that.

> 
>> 3. I thought I understood that UNC path names have now changed to the 
>> generic \\afs\cellname\... rather than the machine-specific 
>> \\%computername%-afs\cellname\... However, I still find only the 
>> latter, despite my "NetbiosName" being "AFS". I installed with all 
>> default settings, except integrated login and my custom cellname and 
>> afsdcell.ini. I am attaching my registry settings for reference. Did I 
>> miss anything? Do I have to make the "afs" hostname resolve to my 
>> loopback address for \\afs to work?
> 
> Yes.  You must install the loopback adapter for this to work.
> 
> If you have the loopback installed and it is not working, please submit 
> a report and
> include your %WINDIR%\TEMP\afsd_init.log file

I think I have the loopback adapter installed. At least I can ping 
127.0.0.1. So I think I'll have to submit that report, too.

> 
>> 4. Adding global drive letters seems to have a syntax inconsistent 
>> with adding user-specific drives, and I haven't been able to make a 
>> global drive actually work. On the other hand, deleting a global drive 
>> seems to shift it to the list of user-specific drives, where I can 
>> neither disable nor delete it.
> 
> This problem is fix in today's test builds.  It only affects people who 
> are not using the loopback adapter.
> 
>> 5. With the old way of mounting root.afs, it was possible to use not 
>> only cell mounts, but also symbolic links in that volume. While I 
>> admit that this has always been a hack, many sites have come to rely 
>> on shortcuts like /afs/@cell, /afs/ThisCell, /afs/z or /afs/zurich in 
>> place of of /afs/zurich.ibm.com.
>>
> In that case do not use Freelance mode.

Where do I turn that off? In general, where are all the new options 
documented? I did not select to locate cells via DNS at install time.

> 
>> 6. If I create a drive mapping for \afs itself, the contents of that 
>> drive seems to be built only dynamically, even though I have not 
>> turned on cell lookups via DNS. Many people expect to be able to 
>> browse through the list of cells specified in afsdcells.ini.
> 
> You cannot browse the list of cells in Freelance mode except those you 
> have mounts.  The client is
> acting as its own afs root instead of relying on the one from the home cell.
> 
>> 7. While I would prefer to use generic \\afs\cellname\... UNC names, 
>> it has proven convenient to also have a global drive mapping for \afs. 
>> Drive mappings provide the only way to CD to an AFS directory. 
>> Unfortunately, the AFS context menu is also only available using drive 
>> letters, not using UNC paths. On the other hand, mapping each cell to 
>> a different drive letter quickly causes a mess and a shortage of drive 
>> letters.
> 
> It is known that the AFS context menu does not work on UNC names.  There 
> is already a feature
> request in the request tracker.
> 
> 
> Jeffrey Altman
> 


-- 
Michael Niksch                     /Zurich/IBM @ IBMCH
IBM Zurich Research Laboratory     nik@zurich.ibm.com
Saeumerstrasse 4                   http://www.zurich.ibm.com/~nik/
CH-8803 Rueschlikon / Switzerland  P: +41-1-724-8913 F: +41-1-724-8080