Searching for Testers was Re: [OpenAFS] OpenAFS site not geared for beginners...

Jason Edgecombe
Sat, 02 Sep 2006 10:25:18 -0400


What if any kind of automated testing tools are used with AFS? In the 
past year, I have been trying to integrate more automated testing into 
my own work and I was wondering what, if any, automated testing is going 
on with openafs?

I fully understand that automated testing can only supplement user 
testing, but I'd like to look into it once I get more into AFS.

Jason Edgecombe

Jeffrey Altman wrote:
> Rodney:
> If there is a problem with a release that you encounter and you don't
> say anything and no one else says anything then it will never be fixed.
> If it is never fixed, then it will become a problem for you when you
> upgrade and it will be too late.
> Even if you don't have time to do a full debug, send a e-mail so
> that at least we know that there is a problem.   Even better would be
> to give your Help Desk personnel a list of instructions to follow when
> they come across a problem so that they can file the bug report to
> and you don't even have to be involved.  If
> the Help Desk is supposed to be supporting OpenAFS for your students
> and staff then that should not necessarily mean that they have to
> contact you whenever there is a problem.
> Simple things such as copy the afsd_init.log file off the machine
> would be quite helpful.  Again, just knowing that there is a problem
> with a given build is helpful.
> Is anyone else experiencing the problems that Rodney has cited?
> Can anyone else help to track them down?
> Jeffrey Altman
> Rodney M Dyer wrote:
>> At 09:14 PM 9/1/2006, Jeffrey Altman wrote:
>>> I feel compelled to point out the fact that I have seen zero bug reports
>>> saying that OpenAFS would not install or that the service would not
>>> start (except on Windows Vista and that is because of bugs in Vista.)
>> Please forgive my knee-jerking and I admit that the problem is not as
>> overblown as I make things out to be.  After being hassled by help desk
>> support, my mood isn't always good.  In the last month I have been
>> called out on as little as two issues that have appeared with the client.
>> 1.  Apparently there is some issue with having the Windows client
>> computername longer than a certain length.
>> 2.  Apparently with the latest build 1.4.2fc2 the client service will
>> crash on startup.
>> I did not have time to perform any analysis of these problems.  I fixed
>> issue 1 just by making the computername shorter.  Issue 2 was fixed
>> after installing 1.4.1 over 1.4.2fc2 - twice.  The first install over
>> 1.4.2fc2 was done by our help desk.  I personally wouldn't have done it
>> that way, thinking that 1.4.1 installer shouldn't know how to uninstall
>> a later version.  And in fact this causes an issue with the AFS client
>> service not being completely removed.  So I needed to re-install 1.4.1
>> one more time, hoping this problem would clear it up, and it did.
>> What we know about issue 2 is that the machine never had AFS on it
>> before.  The machine was a brand new laptop with Windows XP SP2 and all
>> patches.  The user simply downloaded1.4.2fc2 and installed it.  Upon
>> reboot the service crashes.  I did not actually see the crash, but was
>> advised of the crash dump by Windows that occurred.  When I got to the
>> computer, the help desk person had already tried to remove 1.4.2fc2 by
>> installing 1.4.1.
>> This year has been a very complicated and exhausting year for me.  I'm a
>> very tired tech worker in the IT industry.  I love AFS and do want to
>> continue to support the OpenAFS group with testing, however I have been
>> too busy with other things lately to help out.  I found that it has just
>> been too time consuming to keep up with every release of OpenAFS.  The
>> time I have allotted for AFS testing and debugging is very small, if any
>> at all.
>> I'm also a little bit nervous with the fact that as AFS has improved
>> (performance and features), it also still retains some visible
>> instabilities; apparently not only in the client, but in the server
>> too.  We have/are seeing strange issues after moving to 1.4.1.  Issues
>> like...
>> 1.  Users apparently can't add their own permissions back after removing
>> them from their home volume.  This worked in the past. (I think because
>> the user was the owner of the volume).  It no longer works.
>> 2.  Volume replications can sometimes stall, or take longer than normal
>> periods of time, even if the replication is on the same server.
>> 3.  Moving a users volume from one server to another apparently causes
>> some issue with some clients not being able to access the volume.
>> 4.  Other anomalies which we got patches for from the OpenAFS group
>> since we upgraded to 1.4.1.
>> I don't have any good data on these issues yet.  I don't have good time
>> to diagnose these problems.  I mention them only as anecdotal problems. 
>> Even if I can't do anything more about getting more information to you I
>> feel you should at least know something is not right.  I guess I could
>> just keep my mouth closed.
>> Btw, now that it is on my mind, there is also a small cosmetic issue
>> with the afscreds tool drive mapper.  After drives are mapped, the drive
>> mapping listing doesn't refresh unless you click to another tab, then
>> back to the drive mapper (not a show stopper in any way).
>> I don't have any good advice for how to proceed with testing.  I perform
>> many tests when I'm ready to roll out a new client.  I haven't been
>> testing lately because we decided not to upgrade the client this
>> summer.  I need to gear up this fall for the next roll out.  So I will
>> probably increase my testing later this month.  Again time is short. 
>> I'm working so hard these days just to stay a float that I've been
>> thinking my service in the IT industry is almost over.  Managing OSs,
>> applications, and new technologies is just too complex these days. 
>> Maybe it's just me, but I don't understand how other people in IT keep
>> up.  Microsoft I blame the most for all my problems.  Microsoft has
>> created an almost unmanageable working environment (OS and
>> applications).  The toll to keep the beast running is just too high. 
>> I'm about to give up and jump ship.
>> Rodney