[OpenAFS] MS Word crashing when opening files, 1.5.27 client

Hans Melgers hans@enem.nl
Tue, 27 Nov 2007 21:43:11 +0100


Thanks Jeffrey.  The older version was in the early stages, we found it 
also happens with 1.5.25.  On some pc's this happens, on others it doesnt.

So the problem is in winword trying to obtain a lock indefinitely.
What do you recommend ? is there any (registry) setting that does this ?

Would this be a solution ? I mean, would winword accept and act on 
changing this setting to 0 ? :

If the CIFS-AFS server is unable to renew the AFS file server locks, 
then it will invalidate the associated file handles.  This is the same 
behavior that an application will experience if it was using a Windows 
File Share and the connection was broken.   Invalidating the file 
handles prevents subsequent data corruption from taking place.
If you wish to disable the acquisition of locks from the file server, 
this can be performed using the EnableServerLocks registry value.

Value: EnableServerLocks
Type: DWORD {0, 1, 2}
Default: 1

Determines whether or not the AFS file server is contacted for

0: never obtain server locks

1: obtain server locks unless the file server says not to

2: always obtain server locks



Jeffrey Altman wrote:
> Your log shows that at 12:56.57 winword.exe attempted the following
> request that was denied due to the file being in use.
>
> Sequence:    669571
> Date & Time:    11/27/2007 12:56:57 PM
> Event Class:    File System
> Operation:    IRP_MJ_CREATE
> Result:    NOT GRANTED
> Path:    P:\2006\06226-06250\06249\Rapportages\rapportage Friesland 15
> november 2000-02.doc
> TID:    336
> Duration:    92.1418212
> Desired Access:    Generic Read/Write
> Disposition:    Open
> Options:    Synchronous IO Non-Alert, Non-Directory File
> Attributes:    N
> ShareMode:    Read, Write
> AllocationSize:    n/a
>
> Winword.exe then attempts to repeat the request and you terminated the
> logging before a result could be determined.
>
> Sequence:    709074
> Date & Time:    11/27/2007 12:58:29 PM
> Event Class:    File System
> Operation:    IRP_MJ_CREATE
> Result:   
> Path:    P:\2006\06226-06250\06249\Rapportages\rapportage Friesland 15
> november 2000-02.doc
> TID:    336
> Duration:   
> Desired Access:    Generic Read
> Disposition:    Open
> Options:    Synchronous IO Non-Alert, Non-Directory File
> Attributes:    N
> ShareMode:    Read
> AllocationSize:    n/a
>
> I do not see in your log any "DISCONNECTED" errors which is what would
> be listed if the CIFS client timed out and terminated the connection to
> the AFS Client Service.  As far as I can tell WinWord is simply trying
> to open a file it can't obtain a lock on fails, and tries again
> indefinitely.
>
> You said previously that earlier versions of OpenAFS worked better for
> you.  What are the differences in the requests performed between the
> current and previous versions of OpenAFS?   You should be able to
> perform this analysis yourself.  Use the Process Monitor Filter option
> to restrict the displayed events to winword.exe, and paths beginning
> with P:\ and \\AFS
>
> Jeffrey Altman
>
>
>
> Hans Melgers wrote:
>   
>> Hi Jeffrey,
>>
>> We just made the log, its quite big (38MB) so please download it from
>> the site mentioned at the bottom of the mail.
>> I hope you can find something.
>>
>> Kind regards,
>> Hans Melgers
>>
>>
>>
>> Jeffrey Altman wrote:
>>     
>>> Hans:
>>>
>>> Again I must point you at the debugging section of the release notes.
>>> If you would gather the necessary data the problem could be examined
>>> based upon the actual requests that are being issued by the software and
>>> the responses that are provided by the AFS client instead of attempting
>>> to guess at the behavior by looking at the entire system (Office XP,
>>> Windows, AFS client, AFS server) as a black box.
>>>
>>> While it is possible the Netware client might be involved in some way I
>>> would find that unlikely.  The application is requesting a lock and the
>>> file server is unable to grant it because another client already has it.
>>>    In this situation, the file server will report EAGAIN to the client
>>> and the AFS client will retry in the hopes of being able to obtain
>>> the lock.
>>> Perhaps the application is timing out the request before the AFS client
>>> returns.  I don't know for sure because you have not obtained the
>>> necessary data.
>>>
>>> Jeffrey Altman
>>>
>>>
>>> ENEM | Hans Melgers wrote:
>>>  
>>>       
>>>> I dont thing the problem is in the afs client. We found this problem
>>>> only occurs when opening a file that already has been opened by
>>>> another user. We upgraded to office 2003 and it takes alot of time
>>>> before word gives any response.
>>>> This only seems to happen on the "original" pc's, when trying from
>>>> our own pc everything works fine, word opens in read-only mode but
>>>> that doesnt happen on these pc's ?
>>>>
>>>> Maybe it has something to do with the fact that they are also
>>>> running the netware client. Are there any known issues with this
>>>> setup ? Should we run afs before netware ? Maybe check some settings ?
>>>>
>>>>
>>>> -----Original Message-----
>>>> From: Jeffrey Altman [mailto:jaltman@secure-endpoints.com] Sent:
>>>> woensdag 14 november 2007 15:43
>>>> To: Hans Melgers
>>>> Cc: openafs-info@openafs.org
>>>> Subject: Re: [OpenAFS] MS Word crashing when opening files, 1.5.27
>>>> client
>>>>
>>>> Hans:
>>>>
>>>> Please follow the debugging section of the OpenAFS for Windows
>>>> Release Notes.  In particular, use the SysInternals' Process Monitor
>>>> to log the requests from Word so that you can determine which
>>>> requests are failing with 1.5.27 that succeed with 1.5.25.
>>>>
>>>> One change that went into 1.5.27 is support for retrying requests
>>>> when the file server responds to the client with EWOULDBLOCK /
>>>> EAGAIN.  In prior releases the client would simply fail the
>>>> request.  In 1.5.27 the client retries the request every two seconds
>>>> up to 45 seconds until the request completes or the timeout is
>>>> reached.  In the trace log it is the
>>>> cm_Analyze() messages will include the error number.  Possible error
>>>> numbers to look for include:
>>>>
>>>>   UAEWOULDBLOCK 49733416L
>>>>   UAEAGAIN      49733386L
>>>>   EAGAIN              11L
>>>>
>>>>
>>>> Jeffrey Altman
>>>>
>>>>
>>>>
>>>>
>>>> Hans Melgers wrote:
>>>>    
>>>>         
>>>>> Hello,
>>>>>
>>>>> We just installed some XP machines with the latest 1.5.27 afs
>>>>> client. On several of them Word crashes badly when trying to open a
>>>>> .doc file on afs.  Most of them running office XP but also seen on
>>>>> office 2003.
>>>>>
>>>>> I run 1.5.25 and havent seen the problem. Could this be a bug in
>>>>> the latest win client ? acl's are ok. Server is 1.4.4 on freebsd
>>>>> 6.2 Anyone else seeing this behaviour ?
>>>>>
>>>>> In Filelog i only see some callback issues:
>>>>>
>>>>> Wed Nov 14 10:56:32 2007 CB: ProbeUuid for 213.46.24.222:17980
>>>>> failed -01 Wed Nov 14 10:57:11 2007 CB: ProbeUuid for
>>>>> 213.46.24.222:17982 failed -01 Wed Nov 14 10:58:08 2007 CB:
>>>>> ProbeUuid for 213.46.24.222:17987 failed -01 Wed Nov 14 10:58:18
>>>>> 2007 CB: ProbeUuid for 213.46.24.222:17976 failed -01 Wed Nov 14
>>>>> 11:01:25 2007 CB: ProbeUuid for 213.46.24.222:17995 failed -01
>>>>>
>>>>>
>>>>> Any help appreciated.
>>>>>
>>>>> Hans
>>>>>       
>>>>>           
>>>> _______________________________________________
>>>> OpenAFS-info mailing list
>>>> OpenAFS-info@openafs.org
>>>> https://lists.openafs.org/mailman/listinfo/openafs-info
>>>>     
>>>>