samba - May 2016 - Eclipse (Java) locking issues after upgrade (3.6.23 -> 4.4.3)

Hello Jeremy,

Sorry for the long delay for responding to your request (just came back 
to the office after long Victoria day weekend).

I'm not sure if it's Ok to send such large attachements to the list.   
So here's a Dropbox public link to download it:

https://www.dropbox.com/s/o2617jtqfbwlyhc/eclipse-java-samba-level10.log.gz?dl=0

All our connections to this share are not all via Samba.   We use it for 
Linux diskless (via kernel, Pam_Mount) and Windows clients.

That's why we use 'kernel oplocks = yes'.  Maybe we don't need
it? Could
it be the source of our problem?

Please don't hesitate to get in touch if you need more info...

Thanks for the help, Luc.

Le 2016-05-20 18:05, Jeremy Allison a écrit :
> On Fri, May 20, 2016 at 12:01:20PM -0400, Luc Lalonde wrote:
>> Caused
>>   by: java.io.IOException: An error occurred while locking file
>>
"/home/user/.eclipse/org.eclipse.platform_4.6.0_1664049636_macosx_cocoa_x86_64/configuration/org.eclipse.osgi/.manager/.fileTableLock":
>>   "Operation not supported". A common reason is that the file
system or
>> Runtime Environment does not support file locking for that location.
>> Please choose a different location, or disable file locking by passing
>> "-Dosgi.locking=none" as a VM argument.
>>
>> 	at
org.eclipse.osgi.internal.location.Locker_JavaNio.lock(Locker_JavaNio.java:49)
>>
>> 	at
org.eclipse.osgi.storagemanager.StorageManager.lock(StorageManager.java:388)
>>
>> 	at
org.eclipse.osgi.storagemanager.StorageManager.open(StorageManager.java:701)
>>
>> 	at
org.eclipse.osgi.storage.Storage.getChildStorageManager(Storage.java:1785)
>>
>> 	at
org.eclipse.osgi.storage.Storage.getInfoInputStream(Storage.java:1802)
>>
>> 	at org.eclipse.osgi.storage.Storage.<init>(Storage.java:128)
>>
>> 	at org.eclipse.osgi.storage.Storage.createStorage(Storage.java:87)
>>
>> 	at
org.eclipse.osgi.internal.framework.EquinoxContainer.<init>(EquinoxContainer.java:66)
>>
>> 	... 10 more
>>
>> mac-mini:MacOS user$ mount
>>
>> /dev/disk0s2 on / (hfs, local, journaled)
>>
>> devfs on /dev (devfs, local, nobrowse)
>>
>> map -hosts on /net (autofs, nosuid, automounted, nobrowse)
>>
>> map auto_home on /home (autofs, automounted, nobrowse)
>>
>> map -fstab on /Network/Servers (autofs, automounted, nobrowse)
>>
>> //user at sambafileserver.foo.bar/user  <http://p302568 at
moe-180.gi.polymtl.ca/p302568>  on /home/user (smbfs, nodev, nosuid,
automounted, nobrowse, mounted by user)
>>
>>
>> The client is a Mac-Mini running OSX Yosemite with Java 8, update92.
>> We have no problems with Windows 10 clients.
>>
>> Here are the pertinent configuration directives active on the Samba
server:
>>
>>     kernel oplocks = yes
>>     strict locking = No
>>     oplocks = yes
>>     level2 oplocks = yes
>>     deadtime = 15
>>     unix extensions = yes
>>     wide links = yes
>>     allow insecure wide links = yes
>>
>> What's strange  is this was working fine with Samba v3.6.23.    We
>> don't have any problems with Windows clients.
> Turn off "kernel oplocks = yes", you don't need this
> if all access is via Samba.
>
> Do you have a debug level 10 log from the server
> showing access requests at the time of the failure ?
-- 
Luc Lalonde, analyste
-----------------------------
Département de génie informatique:
École polytechnique de MTL
(514) 340-4711 x5049
Luc.Lalonde at polymtl.ca
-----------------------------

On Tue, May 24, 2016 at 12:41:32PM -0400, Luc Lalonde
wrote:
> Hello Jeremy,
> 
> Sorry for the long delay for responding to your request (just came
> back to the office after long Victoria day weekend).
> 
> I'm not sure if it's Ok to send such large attachements to the
list.
> So here's a Dropbox public link to download it:
> 
>
https://www.dropbox.com/s/o2617jtqfbwlyhc/eclipse-java-samba-level10.log.gz?dl=0
> 
> All our connections to this share are not all via Samba.   We use it
> for Linux diskless (via kernel, Pam_Mount) and Windows clients.
> 
> That's why we use 'kernel oplocks = yes'.  Maybe we don't
need it?
> Could it be the source of our problem?
> 
> Please don't hesitate to get in touch if you need more info...
Start by removing all access other than Samba, and see if
you can still reproduce the problem. If not, then it's
the cross-protocol issue.

If you can still reproduce with only SMB clients, ping
me again :-).

Cheers,

	Jeremy.

Hello Jeremy,

Could you send me some references to the ‘cross-protocol issue’ (Bugzilla,
documentation, etc)?

I’ll disable the ‘kernel oplocks’ and see what happens.

We’re getting all sorts of weirdness from our Mac SMB clients since we upgraded
to Samba v4.4.3.

In particular, when we open a session with an OSX (Yosemite) client with SMB
autommount directories, we see multiple SMBD processes that take up almost 100%
of the Samba server’s processes.

I have to kill all of the SMBD processes on the server and reboot the client to
get out of this situation!

If I can’t resolve this before classes begin in September, I’ll have to go back
to using AFP (Netatalk) for our Mac labs.

My Windows 8.1 and Windows 10 clients work fine and do not cause this problem. 

Best regards, Luc.
> On May 27, 2016, at 7:15 PM, Jeremy Allison <jra at samba.org> wrote:
> 
> On Tue, May 24, 2016 at 12:41:32PM -0400, Luc Lalonde wrote:
>> Hello Jeremy,
>> 
>> Sorry for the long delay for responding to your request (just came
>> back to the office after long Victoria day weekend).
>> 
>> I'm not sure if it's Ok to send such large attachements to the
list.
>> So here's a Dropbox public link to download it:
>> 
>>
https://www.dropbox.com/s/o2617jtqfbwlyhc/eclipse-java-samba-level10.log.gz?dl=0
>> 
>> All our connections to this share are not all via Samba.   We use it
>> for Linux diskless (via kernel, Pam_Mount) and Windows clients.
>> 
>> That's why we use 'kernel oplocks = yes'.  Maybe we
don't need it?
>> Could it be the source of our problem?
>> 
>> Please don't hesitate to get in touch if you need more info...
> 
> Start by removing all access other than Samba, and see if
> you can still reproduce the problem. If not, then it's
> the cross-protocol issue.
> 
> If you can still reproduce with only SMB clients, ping
> me again :-).
> 
> Cheers,
> 
> 	Jeremy.