Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   "Benjamin Coddington" <bcodding@redhat.com>
To:     "jie wang" <yjxxtd12@gmail.com>
Cc:     linux-nfs@vger.kernel.org
Subject: Re: Question abount sm-notify when use NFSv3 lock
Date:   Wed, 06 Jul 2022 10:39:41 -0400
Message-ID: <3D87B9ED-3A00-478B-AC17-435B71D0A349@redhat.com>
In-Reply-To: <CACt_J9PHSjkz_-x0K=7+AYjiX1Ur5cV+brC9Tv4i7dkG=NSBuQ@mail.gmail.com>
References: <CACt_J9PHSjkz_-x0K=7+AYjiX1Ur5cV+brC9Tv4i7dkG=NSBuQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

On 6 Jul 2022, at 8:28, jie wang wrote:

> Hi, all
>   When we use NFSv3, we have a LoadBalance in front of NFS server. For
> example, LoadBalance's ip is ip2 and NFS server's ip is ip1, and
> client use ip2 to mount.
>
>   Now client use flock to lock file, then I restart NFS server and
> execute sm-notify -f. Then the problem occurs, the sm-notify request's
> src ip is ip1, not ip2, so rpc.statd will ignore this notify, because
> it does not match ip2 when mount, so client will not reclaim lock, and
> lock lost when restart NFS server.
>
>   Do you know how to address this ? Thanks a lot.

The sm-notify(8) man page shows you can use '-v' to specify an ipaddr or
hostname.

Ben