[Gluster-devel] How to find total number of gluster mounts?

[Niels de Vos]

On Wed, Jun 03, 2015 at 11:37:58AM +0530, Raghavendra Talur wrote:
> On Wednesday 03 June 2015 09:13 AM, Pranith Kumar Karampuri wrote:
> >
> >
> >On 06/01/2015 11:07 AM, Bipin Kunal wrote:
> >> Hi All,
> >>
> >>   Is there a way to find total number of gluster mounts?
> >>
> >>   If not, what would be the complexity for this RFE?
> >>
> >>   As far as I understand finding the number of fuse mount should be
> >> possible but seems unfeasible for nfs and samba mounts.
> >True. Bricks have connections from each of the clients. Each of
> >fuse/nfs/glustershd/quotad/glfsapi-based-clients(samba/glfsheal) would
> >have separate client-context set on the bricks. So We can get this
> >information. But like you said I am not sure how it can be done in nfs
> >server/samba. Adding more people.
> 
> Depends on why you would want to know about the clients:
> 
> 1. For most of the use cases, admin might just need to know how many
> Samba/NFS servers are currently using the given volume(Say just to perform umount everywhere).
> In this case, each Samba/NFS server is just like a FUSE mount and we can use the same
> technique that we would use for the above case that Pranith has mentioned.
> 
> 2. If the requirement is to identify all the machines which are accessing a volume,
> (probable use case:- you may want a end-user to close a file etc)
> above method won't be sufficient. To get details of SMB clients, you would have to
> run 'smbstatus' command on all SMB server nodes and it would output details of
> connected SMB clients in this format.
> PID     Username      Group         Machine            Protocol Version     Service      pid     machine       Connected at

Gluster/NFS supports the 'showmount' command (over the MOUNT RPC
protocol). It can be used to list all the NFS-clients that have a
volume/subdir mounted.

This list should not be 100% trusted though. NFSv3 uses the MOUNT RPC
protocol to get the file-handle for the mountpoint. After that, the
NFSv3 protocol can use the export until it wants to. When the NFS-client
unmounts the export, it sends the UMNT procedure to the NFS-server what
causes the NFS-client/export combination to be removed from the
client-list (showmount output).

A client that does not send a UMNT, will not be removed from the list of
active clients. This can happen when a client does a umount during
network issues, or a client spontaneously reboots (or kernel panic or
..). Very similar are clients that mount the exact same export/subdir in
multiple mountpoints. The NFS-server can not differentiate between a
client that did not set a UMNT, or a client that mounts the same
export/subdir more than once. These clients will only be listed once.

HTH,
Niels