Issue 1 in tsumufs: Ping NFS server

5 views
Skip to first unread message

codesite...@google.com

unread,
Jun 4, 2009, 1:22:42 PM6/4/09
to google-...@googlegroups.com
Updates:
Status: Accepted

Comment #1 on issue 1 by june.tate: Ping NFS server
http://code.google.com/p/tsumufs/issues/detail?id=1

Hm. I would agree, but the problem is that it calls out to yet another
shell command.
Since we're running multithreaded python, using popen(2) in this way is very
dangerous for the life of the program -- especially when we call it
multiple times.
Further, rpcinfo is VERY NFS specific, so I'm pretty reluctant to integrate
it into
the codebase. Really, we should setup an abstraction class to move the NFS
specifics
outside of the syncthread entirely and allow for using alternate classes for
differing FSes.

As a result, I would like to see the RPC protocol implemented in python as
an
extension instead, or at least a simple RPC implementation in pure python.
TBH, the
/only/ reason why we shell out to the mount command to setup NFS in the
first place
is because the mount syscall isn't portable and isn't even really available
to
userspace without crazy pretzel-like contortions.

I'm going to mark this as Accepted until one of the two above problems can
be
integrated somehow.

--
You received this message because you are listed in the owner
or CC fields of this issue, or because you starred this issue.
You may adjust your issue notification preferences at:
http://code.google.com/hosting/settings

megabast

unread,
Jun 15, 2009, 12:25:07 PM6/15/09
to google-tsumufs
Hi,
I've found a new solution for checking nfs server availability.
I've used pynfs (http://freshmeat.net/projects/pynfs/) which works
with nfs v4.
It's an extern module and it's very simple to use it. So it's a new
dependence for tsumufs with the RPC protocol implementation in pure
python.
Waiting for your feelings about it.
Bastien

On Jun 4, 7:22 pm, codesite-nore...@google.com wrote:
> Updates:
>         Status: Accepted
>
> Comment #1 on issue 1 by june.tate: Ping NFS serverhttp://code.google.com/p/tsumufs/issues/detail?id=1

June Tate-Gans

unread,
Jun 17, 2009, 4:15:31 PM6/17/09
to google-...@googlegroups.com
This looks VERY promising for what the bug needs. My only concern is
that it seems to be NFSv4 only, while the primary NFS implementations
out there are v3 (from past experience, most v4 clients and servers
are horribly bug ridden and still haven't been deployed very heavily).
Perhaps we should extract or use the RPC implementation and just use
that to verify some checks in the ping method.

Great find! I can't look into it much more right now, but I'll
definitely be having a closer look sometime tonight or tomorrow.
Thanks!
--
June Tate-Gans | Don't try to outweird me, three-eyes. I get stranger things
Sysops, SEA | than you free with my breakfast cereal. --Zaphod Beeblebrox

codesite...@google.com

unread,
Jun 17, 2009, 4:17:28 PM6/17/09
to google-...@googlegroups.com
Updates:
Owner: june.tate

Comment #2 on issue 1 by june.tate: Ping NFS server
http://code.google.com/p/tsumufs/issues/detail?id=1

On Jun 15th, megabast wrote:
> Hi,
> I've found a new solution for checking nfs server availability.
> I've used pynfs (http://freshmeat.net/projects/pynfs/) which works
> with nfs v4.
> It's an extern module and it's very simple to use it. So it's a new
> dependence for tsumufs with the RPC protocol implementation in pure
> python.

This looks VERY promising for what the bug needs. My only concern is


that it seems to be NFSv4 only, while the primary NFS implementations
out there are v3 (from past experience, most v4 clients and servers
are horribly bug ridden and still haven't been deployed very heavily).
Perhaps we should extract or use the RPC implementation and just use
that to verify some checks in the ping method.

Great find! I can't look into it much more right now, but I'll
definitely be having a closer look sometime tonight or tomorrow.
Thanks!

--

Reply all
Reply to author
Forward
0 new messages