Help: NFS/mount params ineffective

Help: NFS/mount params ineffective

Post by mt.. » Sat, 08 May 1999 04:00:00



Hi,


: I've used several different sets of parameters and the results have always
: been the same... I get an RPC timeout after 2 minutes and 3 seconds.  Here
: are a couple of examples that I've tried:
[...]

: Now, 'rhost' is shutdown right now.  I've been sniffing the wire as I do this
: and here's what happens:  I send out a billion arp requests and since none of
: them get answered... it quits.        I have the feeling this is an RPC thing.  It
: seems no matter what, I can't get RPC to wait longer (than 2min03secs) before
: it times out... nor, can I force it to timeout sooner (as I thought the last
: example above would do).

: Can someone help?  I'm sure this can't be that difficult, I'm just overlooking
: something.

like booting 'rhost'? If I understood correctly, you are trying
to mount a filesystem from 'rhost', which is not running, and
thus can't possibly respond to RPC requests, causing a timeout.
Fire up 'rhost' and you should be set, as long as it exports
the right filesystem!

Bye,
Mike

--
<< the above email addr might disappear, reply to: >>
<<         Michael.Sievers -(at)- desy.de          >>

Black holes are where God divided by zero.

 
 
 

Help: NFS/mount params ineffective

Post by deu.. » Tue, 11 May 1999 04:00:00


You are correct, I am trying to mount a machine that is not running.
But, I am doing it so that I may debug my NFS client.  If I understand
NFS correctly, which I probably don't, you may feed different params to
the mount command and affect its behavior.  The behavior I am trying to
affect is the timeout behavior.  I want to see some sign that I am
successfully doing this.  So, I tried to mount a dead machine and feed
it parameters that would make my client hang for a long time, trying to
mount it.  And, I fed it parameters that would make my client quit
trying to mount immediately.

NFS shouldn't care that the machine is dead.  As far as NFS knows, the
server is just being bombarded and has no time to respond.  But, no
matter what I tried to feed the mount command, my client side RPC always
timed out after two minutes.

What I'm guessing is happening is that since RPC is not even connecting
on the remote machine (as it would even if the NFS server was swamped)
that it is giving up.  If this is correct, then I need to know how to
affect my client-side RPC... ????  From what I understand, NFS is
supposed to force RPC to contiue to try to fulfill its requests, if you
try mounting with the right parameters.

What gives?

Thanks,

Bob



> Hi,


> : I've used several different sets of parameters and the results have
always
> : been the same... I get an RPC timeout after 2 minutes and 3
seconds.  Here
> : are a couple of examples that I've tried:
> [...]

> : Now, 'rhost' is shutdown right now.  I've been sniffing the wire as
I do this
> : and here's what happens:  I send out a billion arp requests and
since none of
> : them get answered... it quits.   I have the feeling this is an RPC
thing.  It
> : seems no matter what, I can't get RPC to wait longer (than
2min03secs) before
> : it times out... nor, can I force it to timeout sooner (as I thought
the last
> : example above would do).

> : Can someone help?  I'm sure this can't be that difficult, I'm just
overlooking
> : something.

> like booting 'rhost'? If I understood correctly, you are trying
> to mount a filesystem from 'rhost', which is not running, and
> thus can't possibly respond to RPC requests, causing a timeout.
> Fire up 'rhost' and you should be set, as long as it exports
> the right filesystem!

> Bye,
> Mike

> --
> << the above email addr might disappear, reply to: >>
> <<         Michael.Sievers -(at)- desy.de          >>

> Black holes are where God divided by zero.

--== Sent via Deja.com http://www.deja.com/ ==--
---Share what you know. Learn what you don't.---

 
 
 

Help: NFS/mount params ineffective

Post by mt.. » Thu, 13 May 1999 04:00:00


Hi,

I think what you want to do is have a working nfs server,
allowing your clients to mount the filesystem and having some
programs accessing files. Then you want to kill the nfs server,
by powering it off or just unplugging the network connection.
The options you give for mounting affect the nfs behaviour in
this case, like retrying until a success is achieved or
giving up and returning an error. I believe that leaving the
nfs server powered off from the start just tells you how long
an rpc timeout is, but killing it once its filesystem is being
accessed gives you the information you want.

Is this what you want to do?

Bye,
Mike


: You are correct, I am trying to mount a machine that is not running.
: But, I am doing it so that I may debug my NFS client.  If I understand
: NFS correctly, which I probably don't, you may feed different params to
: the mount command and affect its behavior.  The behavior I am trying to
: affect is the timeout behavior.  I want to see some sign that I am
: successfully doing this.  So, I tried to mount a dead machine and feed
: it parameters that would make my client hang for a long time, trying to
: mount it.  And, I fed it parameters that would make my client quit
: trying to mount immediately.

: NFS shouldn't care that the machine is dead.  As far as NFS knows, the
: server is just being bombarded and has no time to respond.  But, no
: matter what I tried to feed the mount command, my client side RPC always
: timed out after two minutes.

: What I'm guessing is happening is that since RPC is not even connecting
: on the remote machine (as it would even if the NFS server was swamped)
: that it is giving up.  If this is correct, then I need to know how to
: affect my client-side RPC... ????  From what I understand, NFS is
: supposed to force RPC to contiue to try to fulfill its requests, if you
: try mounting with the right parameters.

: What gives?

: Thanks,

: Bob



:> Hi,
:>

:> : I've used several different sets of parameters and the results have
: always
:> : been the same... I get an RPC timeout after 2 minutes and 3
: seconds.  Here
:> : are a couple of examples that I've tried:
:> [...]
:>
:> : Now, 'rhost' is shutdown right now.  I've been sniffing the wire as
: I do this
:> : and here's what happens:  I send out a billion arp requests and
: since none of
:> : them get answered... it quits.  I have the feeling this is an RPC
: thing.  It
:> : seems no matter what, I can't get RPC to wait longer (than
: 2min03secs) before
:> : it times out... nor, can I force it to timeout sooner (as I thought
: the last
:> : example above would do).
:>
:> : Can someone help?  I'm sure this can't be that difficult, I'm just
: overlooking
:> : something.
:>
:> like booting 'rhost'? If I understood correctly, you are trying
:> to mount a filesystem from 'rhost', which is not running, and
:> thus can't possibly respond to RPC requests, causing a timeout.
:> Fire up 'rhost' and you should be set, as long as it exports
:> the right filesystem!
:>
:> Bye,
:> Mike
:>
:> --
:> << the above email addr might disappear, reply to: >>
:> <<         Michael.Sievers -(at)- desy.de          >>
:>
:> Black holes are where God divided by zero.
:>

: --== Sent via Deja.com http://www.deja.com/ ==--
: ---Share what you know. Learn what you don't.---

--
<< the above email addr might disappear, reply to: >>
<<         Michael.Sievers -(at)- desy.de          >>

Black holes are where God divided by zero.

 
 
 

Help: NFS/mount params ineffective

Post by deu.. » Thu, 13 May 1999 04:00:00


Hey again,

Well, that makes sense.  But, what if both machines go down at the
same time (power failure)...  and furthermore, what if the NFS client
comes back to life much quicker than the NFS server (because it has a
smaller disk drive and e2fsck takes much less time).... then,
effectively it appears to be the same as what I'm doing:  one machine
trying to mount another machine (when the other machine isn't even
running yet).

Right now, we are having problems with this on our network.  When we
have a power failure, at least one of our client machines (every time)
does not mount the server due to timeout.  What we want to do is to set
up our NFS clients to try and try forever and ever, until they are
connected to the server.  So, I was trying to manipulate the timeout
period and force RPC to not give up so soon....

???  Maybe I cannot do what I'm trying to do?

Bob

> I think what you want to do is have a working nfs server,
> allowing your clients to mount the filesystem and having some
> programs accessing files. Then you want to kill the nfs server,
> by powering it off or just unplugging the network connection.
> The options you give for mounting affect the nfs behaviour in
> this case, like retrying until a success is achieved or
> giving up and returning an error. I believe that leaving the
> nfs server powered off from the start just tells you how long
> an rpc timeout is, but killing it once its filesystem is being
> accessed gives you the information you want.

> Is this what you want to do?

> Bye,
> Mike


> : You are correct, I am trying to mount a machine that is not running.
> : But, I am doing it so that I may debug my NFS client.  If I
understand
> : NFS correctly, which I probably don't, you may feed different params
to
> : the mount command and affect its behavior.  The behavior I am trying
to
> : affect is the timeout behavior.  I want to see some sign that I am
> : successfully doing this.  So, I tried to mount a dead machine and
feed
> : it parameters that would make my client hang for a long time, trying
to
> : mount it.  And, I fed it parameters that would make my client quit
> : trying to mount immediately.

> : NFS shouldn't care that the machine is dead.  As far as NFS knows,
the
> : server is just being bombarded and has no time to respond.  But, no
> : matter what I tried to feed the mount command, my client side RPC
always
> : timed out after two minutes.

> : What I'm guessing is happening is that since RPC is not even
connecting
> : on the remote machine (as it would even if the NFS server was
swamped)
> : that it is giving up.  If this is correct, then I need to know how
to
> : affect my client-side RPC... ????  From what I understand, NFS is
> : supposed to force RPC to contiue to try to fulfill its requests, if
you
> : try mounting with the right parameters.

> : What gives?

> : Thanks,

> : Bob



> :> Hi,
> :>

> :> : I've used several different sets of parameters and the results
have
> : always
> :> : been the same... I get an RPC timeout after 2 minutes and 3
> : seconds.  Here
> :> : are a couple of examples that I've tried:
> :> [...]
> :>
> :> : Now, 'rhost' is shutdown right now.  I've been sniffing the wire
as
> : I do this
> :> : and here's what happens:  I send out a billion arp requests and
> : since none of
> :> : them get answered... it quits.     I have the feeling this is an RPC
> : thing.  It
> :> : seems no matter what, I can't get RPC to wait longer (than
> : 2min03secs) before
> :> : it times out... nor, can I force it to timeout sooner (as I
thought
> : the last
> :> : example above would do).
> :>
> :> : Can someone help?  I'm sure this can't be that difficult, I'm
just
> : overlooking
> :> : something.
> :>
> :> like booting 'rhost'? If I understood correctly, you are trying
> :> to mount a filesystem from 'rhost', which is not running, and
> :> thus can't possibly respond to RPC requests, causing a timeout.
> :> Fire up 'rhost' and you should be set, as long as it exports
> :> the right filesystem!
> :>
> :> Bye,
> :> Mike
> :>
> :> --
> :> << the above email addr might disappear, reply to: >>
> :> <<         Michael.Sievers -(at)- desy.de          >>
> :>
> :> Black holes are where God divided by zero.
> :>

> : --== Sent via Deja.com http://www.deja.com/ ==--
> : ---Share what you know. Learn what you don't.---

> --
> << the above email addr might disappear, reply to: >>
> <<         Michael.Sievers -(at)- desy.de          >>

> Black holes are where God divided by zero.

--== Sent via Deja.com http://www.deja.com/ ==--
---Share what you know. Learn what you don't.---
 
 
 

Help: NFS/mount params ineffective

Post by Stefan Monnie » Thu, 13 May 1999 04:00:00



> connected to the server.  So, I was trying to manipulate the timeout
> period and force RPC to not give up so soon....

My understanding is that the options you're using apply the NFS protocol
while the problem you're talking about apply to the MOUNT protocol.

Quote:> ???  Maybe I cannot do what I'm trying to do?

You can of course do it.
Looking at the Solaris manpage and hoping that Linux uses the same
names for the same functionality:

          bg|fg          If the first attempt fails, retry in the
                         background,  or, in the foreground.  The
                         default is fg.

          retry=n        The number of times to retry  the  mount
                         operation.  The default is 10000.

maybe retry=0 will loop for ever ?
But even if this doesn't work, you can still do it with something like:

        until mount <mntopts>; do sleep 10; done

Finally, as long as your boot-up scripts don't depend on the NFS server being
up, I'd recommend the use of an automounter since it will then try the mount
each time the mount point is accessed.

        Stefan

 
 
 

1. web server mount files over nfs,aix net params tuning

Our web server (aix 4.3.3) has its web pages (basically 2
filesystems nfs mounted) on another rs6k (aix 4.2.1)
Now is there any network parameter to be changed from its
system defaults (and the changing of no of nfsd , biod necessary)
, for a very average (i mean typical) web access load, so there
will not be any access hiccups/errors.
And if so, do these params have to be done on nfs file server or/and
client ?
Pls advise with your experience.

2. pcAnywhere connect problems

3. kernel compiled with NFS mount params?

4. HELP: WRONG CPU DETECTION

5. /var/mail: to NFS mount or not to NFS mount?

6. linux driver wtih STREAMS system

7. nfs mount to a nfs mount fails

8. NetBSD + Perl 5.001 problem

9. nfs mount problem: mount: can't get address for nfs-server.kauai

10. HELP: NFS MOUNT through mount( ) syscall

11. HELP NFS permission denied on NFS mount point

12. NFS mount won't mount at boot, but mount -a works fine.

13. PARAM 3/4: PARAM() support in modules.