Gluster users - Nov 2017 - SMB copies failing with GlusterFS 3.10

Hi all

We've got a brand new 6-node GlusterFS 3.10 deployment (previous 20 nodes
were GlusterFS 3.6). Running on CentOS 7 using legit repos, so
glusterfs-3.10.7-1.el7.x86_64 is the base.

Our issue is that when we create a file with a Gluster client, e.g. a Mac
or Windows machine, it works fine. However if we copy a file from a Mac or
Windows machine to the Samba share, it fails with a complaint that the file
already exists, even though it doesn't (or didn't). It appears as though
the file is tested to see if it exists, and it doesn't, so then the client
goes and tries to create the file but at that stage it DOES exist, maybe
something to do with the previous stat? Anyway, it is repeatable and
killing us! NFS clients work fine on any platform, but SMB does not. There
aren't that many client-side options for SMB mounts so the solution has to
be server-side.

Here is a pcap of the copy attempt from one computer:

https://www.dropbox.com/s/yhn3s1qbxtdvnoh/sambacap.pcapng?dl=0

You'll see a the request to look for the file which results in a
STATUS_OBJECT_NAME_NOT_FOUND (good), followed by a STATUS_SHARING_VIOLATION
(???) followed by a STATUS_OBJECT_NAME_COLLISION (bad).

Here are the options from the volume:

Options Reconfigured:

nfs.acl: off

features.cache-invalidation: on

storage.batch-fsync-delay-usec: 0

transport.address-family: inet6

nfs.disable: on

performance.stat-prefetch: off

server.allow-insecure: on

ganesha.enable: ganesha-nfs

user.smb: enable

Any thoughts on why samba isn't enjoying our copies?

Thanks!

Brett.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.gluster.org/pipermail/gluster-users/attachments/20171114/9a49091e/attachment.html>

What is the volume configuration, is it replicated, distributed,
distribute-replicate, disperse?

have you tried setting:
performance.strict-write-ordering to on?

On 14 November 2017 at 06:24, Brett Randall <brett.randall at gmail.com>
wrote:
> Hi all
>
> We've got a brand new 6-node GlusterFS 3.10 deployment (previous 20
nodes
> were GlusterFS 3.6). Running on CentOS 7 using legit repos, so
> glusterfs-3.10.7-1.el7.x86_64 is the base.
>
> Our issue is that when we create a file with a Gluster client, e.g. a Mac
> or Windows machine, it works fine. However if we copy a file from a Mac or
> Windows machine to the Samba share, it fails with a complaint that the file
> already exists, even though it doesn't (or didn't). It appears as
though
> the file is tested to see if it exists, and it doesn't, so then the
client
> goes and tries to create the file but at that stage it DOES exist, maybe
> something to do with the previous stat? Anyway, it is repeatable and
> killing us! NFS clients work fine on any platform, but SMB does not. There
> aren't that many client-side options for SMB mounts so the solution has
to
> be server-side.
>
> Here is a pcap of the copy attempt from one computer:
>
> https://www.dropbox.com/s/yhn3s1qbxtdvnoh/sambacap.pcapng?dl=0
>
> You'll see a the request to look for the file which results in a
> STATUS_OBJECT_NAME_NOT_FOUND (good), followed by a STATUS_SHARING_VIOLATION
> (???) followed by a STATUS_OBJECT_NAME_COLLISION (bad).
>
> Here are the options from the volume:
>
> Options Reconfigured:
>
> nfs.acl: off
>
> features.cache-invalidation: on
>
> storage.batch-fsync-delay-usec: 0
>
> transport.address-family: inet6
>
> nfs.disable: on
>
> performance.stat-prefetch: off
>
> server.allow-insecure: on
>
> ganesha.enable: ganesha-nfs
>
> user.smb: enable
>
> Any thoughts on why samba isn't enjoying our copies?
>
> Thanks!
>
> Brett.
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://lists.gluster.org/mailman/listinfo/gluster-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.gluster.org/pipermail/gluster-users/attachments/20171128/e102c10c/attachment.html>