Re: [Tracker] Disable indexing on NFS

[Martyn Russell <martyn lanedo com>]

On 16/09/13 12:03, Philip Van Hoof wrote:
> Hi Janinko,

Hi all,

> I don't think tracker's fs miner detects NFS mount points (yet, as I
> think a patch for that would be a great idea). You could add the mount
> point to your list of ignores in Tracker's config as a temporary solution.

No we don't sadly. The only option the crawler really has with file 
enumeration is to not follow symlinks.

However, there is a way to check if a mount point is an NFS mount with 
GLib using the:

1. G_VOLUME_IDENTIFIER_KIND_NFS_MOUNT and g_volume_get_identifier() API.

or

2. if (strcmp (g_unix_mount_point_get_fs_type (mountpoint), "nfs") == 0) 
approach IIRC.

--

This sort of thing should be an addition to the 
src/libtracker-miner/tracker-storage.[ch] ideally and we should probably 
have an option in the preferences to not crawl "off" our immediate file 
system, so we can do this more generally and not just for NFS.

> ps. If writing a patch to detect NFS mount points, the patch shouldn't
> make it impossible to do so. As scanning NFS mount points ought to be a
> possible use-case.

The question is really whether we consider ANY mount point. Currently we 
index all mount points separately but we also coalesce them, e.g. if we 
have:

  /media/a
  /media/b

both should be indexed. However, if we have:

  /media/a
  /media/a/nfs-mounted-location/

The second of the two should be consolidated into the first because it's 
a parent of the second.

That's all fine - but what I think is happening here is we are indexing 
$HOME and there is an nfs mount in $HOME/foo perhaps? It's also quite 
possible that people mount other file system types in $HOME/bar, etc. So 
the question becomes: do we want to have an option to stay on the same 
file system or not - that should cover all NFS and other file system 
type cases.

Thoughts?

--
Regards,
Martyn

Founder & Director @ Lanedo GmbH.
http://www.linkedin.com/in/martynrussell