Tunable timeout in cifs module.

Tunable timeout in cifs module.

[Sachin Prabhu]

Hello Steve/linux-cifs,

We have received a request to make the timeout in the cifs module
tunable.

I have gone through the code to understand timeouts in the cifs client.
http://sprabhu.blogspot.in/2015/08/investigation-into-effects-of-server
.html

The way to allow user settable timeout values is to make the value of
 SMB_ECHO_INTERVAL tunable. Do you think this is a good idea?

Sachin Prabhu

Re: Tunable timeout in cifs module.

[Steve French]

On Tue, Oct 20, 2015 at 3:20 AM, Sachin Prabhu <sprabhu-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> wrote:
> Hello Steve/linux-cifs,
>
> We have received a request to make the timeout in the cifs module
> tunable.
>
> I have gone through the code to understand timeouts in the cifs client.
> http://sprabhu.blogspot.in/2015/08/investigation-into-effects-of-server
> .html
>
> The way to allow user settable timeout values is to make the value of
>  SMB_ECHO_INTERVAL tunable. Do you think this is a good idea?

Seems like a reasonable idea - although we might want to treat this
differently for the case (longer timeouts? or different retry
behavior?) for the case of CA shares (or mounts with persistent handle
or resilient handle mount parms - the patches that I sent out a few
weeks ago). We also need to clean up reconnect behavior (byte range
lock recovery) and make sure for the persistent handle case (and
presumably resilient handles too) that we reopen all files on the
mount on reconnect.



-- 
Thanks,

Steve

Re: Tunable timeout in cifs module.

[Steve French]

This reconnect discussion should probably be broadened to see how it
would change (the retry/timeout/reconnect) when we are tied into the
(samba, user space) Witness protocol client that Guenther and others
have been experimenting with.  Perhaps ioctls down to cifs.ko from the
witness protocol client can help us in failover more rapidly.
Similarly if we can try different reconnect strategies (if this share
is a dfs referral then we may already have replicas we can quickly
connect to).

See http://www.sambaxp.org/archive_data/SambaXP2015-SLIDES/wed/track1/sambaxp2015-wed-track1-Guenther_Deschner-ImplementingTheWitnessProtocolInSamba.pdf
for some early description of Guenther's work.

On Tue, Oct 20, 2015 at 12:21 PM, Steve French <smfrench-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> On Tue, Oct 20, 2015 at 3:20 AM, Sachin Prabhu <sprabhu-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> wrote:
>> Hello Steve/linux-cifs,
>>
>> We have received a request to make the timeout in the cifs module
>> tunable.
>>
>> I have gone through the code to understand timeouts in the cifs client.
>> http://sprabhu.blogspot.in/2015/08/investigation-into-effects-of-server
>> .html
>>
>> The way to allow user settable timeout values is to make the value of
>>  SMB_ECHO_INTERVAL tunable. Do you think this is a good idea?
>
> Seems like a reasonable idea - although we might want to treat this
> differently for the case (longer timeouts? or different retry
> behavior?) for the case of CA shares (or mounts with persistent handle
> or resilient handle mount parms - the patches that I sent out a few
> weeks ago). We also need to clean up reconnect behavior (byte range
> lock recovery) and make sure for the persistent handle case (and
> presumably resilient handles too) that we reopen all files on the
> mount on reconnect.
>
>
>
> --
> Thanks,
>
> Steve



-- 
Thanks,

Steve