NFS Problems
Vicky Tanya Seno
Does anyone know what this error means ??
NFS getattr failed for server nfs1: error 27 (RPC: Received
disconnect from remote)
We have a new HP machine running HPUX 11i that exports the students
home directories to all types of sun workstation in the labs. Now all
the SUN keep getting the above error message.
Before we had another HP box running 11 export the home directories
and there were no problems.
Can someone help me out here ? ?
THANK YOU!!
Tanya
--
Vicky "Tanya" Seno
UNIX Administrator
Information Systems Group
College of Engineering and Computer Science
California State University, Northridge
18111 Nordhoff Street
Northridge, CA 91330-8295
Work: 818.677.3919
Fax: 818.677.4192
Pager: 818.372.4688
dmc
Sorry I am not really a HPUX guy but more Solaris, anyway I'll give it a
shot.
With this error, it seems like the export is not happening at all, right? So
in other word machine HP need to write to machine SUN? right? If yes,
proceed to the next line.
The new HP hostname, have you added this on all the SUN's .rhosts or
hosts.equiv file? Simple test, from HP try doing a "rcp text.txt
SUN_hostname:/tmp" . If allowed you'll be able to see the test.txt file on
SUN's /tmp directory, i.e., permission is OK if NOT then that is the
problem.
HTH
DMC
"Vicky Tanya Seno" <da...@ecs.csun.edu> wrote in message
news:3CBDB520...@ecs.csun.edu...
Vicky Tanya Seno
Thanks for the help but the way i do is by using /etc/exportfs on the HPUX
machine (It's like /etc/dfs/dfstab on suns). Then on the SUNs, I use either auto
mounting /etc/auto_master & /etc/auto_direct or just specify the mount
/etc/vfstab.
I don't use .rhosts at all.
Tanya
David Robinson
> Does anyone know what this error means ??
> NFS getattr failed for server nfs1: error 27 (RPC: Received
> disconnect from remote)
It means the server is unilaterally breaking the TCP
connection with the client. This is a normal thing
for a server to do.
> Can someone help me out here ? ?
Stop mounting the server with the "soft" mount option, then
these recoverable "errors" will be recovered.
This is not unique to HP-UX, Solaris servers will also
unilaterally disconnect clients.
-David
Vicky Tanya Seno
the moment most the mount are TCP and its generating to much traffic.
nfsstat -m
/home/users1 from nfs1:/home/users1
Flags:
vers=3,proto=tcp,sec=sys,soft,intr,link,symlink,rsize=32768,wsize=32768,retrans=5
/home/users2 from nfs1:/home/users2
Flags:
vers=3,proto=tcp,sec=sys,soft,intr,link,symlink,rsize=32768,wsize=32768,retrans=5
/opt/.source from nfs1:/opt/.source
Flags:
vers=3,proto=tcp,sec=sys,soft,intr,link,symlink,acl,rsize=32768,wsize=32768,retrans=5
/opt/cecs from nfs1:/tmp_mnt/sun
Flags:
vers=3,proto=tcp,sec=sys,soft,intr,link,symlink,acl,rsize=32768,wsize=32768,retrans=5
/var/mail from nfs2:/var/mail
Flags:
vers=3,proto=udp,sec=sys,soft,intr,link,symlink,acl,rsize=32768,wsize=32768,retrans=5
Lookups: srtt=119 (297ms), dev=103 (515ms), cur=66 (1320ms)
#
Tanya
David Robinson wrote:
--
Benny
>
> Is there anyway to force Solaris to use UDP for NFS mounts. Because at
> the moment most the mount are TCP and its generating to much traffic.
>
> nfsstat -m
> /home/users1 from nfs1:/home/users1
> Flags:
> vers=3,proto=tcp,sec=sys,soft,intr,link,symlink,rsize=32768,wsize=32768,retrans=5
From 'man mount_nfs' :
Hard versus Soft
File systems that are mounted read-write or that con-
tain executable files should always be mounted with
the hard option. Applications using soft mounted file
systems may incur unexpected I/O errors, file corrup-
tion, and unexpected program core dumps. The soft
option is not recommended.
proto=<netid>
<netid> is a value of network_id field from
entry in the /etc/netconfig file. By default,
the transport protocol used for the NFS mount
will be first available connection oriented
transport supported on both the client and the
server. If no connection oriented transport is
found, then the first available connectionless
transport is used. This default behavior can be
overridden with the proto=<netid> option.
UDP is not used anymore for NFS for a reason but ... it is up to you to
decide.
regards,
Benny
--
Disclaimer:
My opinions are not to be confused with those of my employer, Sun
Microsystems.
As in "a cat is not to be confused with a pack of lions". Purrrrrrrrrr.
Darren Dunham
> In article <3CBDCEBA...@ecs.csun.edu>,
> Vicky Tanya Seno <da...@ecs.csun.edu> wrote:
>> Is there anyway to force Solaris to use UDP for NFS mounts. Because at
>> the moment most the mount are TCP and its generating to much traffic.
*blink* Huh? Why do you think that TCP generates much more than UDP
for NFS mounts?
> Switch to NFS v2 in the mount.
NFS v2/v3 and UDP/TCP are not related. You can run any of the four
combinations.
UDP/2
UDP/3
TCP/2
TCP/3
The option to change the transport is covered in this section of the man
page.
proto=<netid>
<netid> is a value of network_id field from
entry in the /etc/netconfig file. By default,
the transport protocol used for the NFS mount
will be first available connection oriented
transport supported on both the client and the
server. If no connection oriented transport is
found, then the first available connectionless
transport is used. This default behavior can be
overridden with the proto=<netid> option.
I don't think it's the most clear section, but when it says 'connection
oriented' versus 'connectionless', it means that it will try 'tcp'
before trying 'udp'. <netid> can be 'tcp' or 'udp' in this case. This
is better shown by the examples below.
Example 3: Mounting An NFS File System Over Version 2, With
The UDP Transport
To mount an NFS file system over Version 2, with the UDP
transport:
example# mount -o vers=2,proto=udp serv:/usr/src /usr/src
--
Darren Dunham ddu...@taos.com
Unix System Administrator Taos - The SysAdmin Company
Got some Dr Pepper? San Francisco, CA bay area
< How are you gentlemen!! Take off every '.SIG'!! >
David Robinson
> Is there anyway to force Solaris to use UDP for NFS mounts. Because at
> the moment most the mount are TCP and its generating to much traffic.
"man mount_nfs"
See the "proto" option.
But switching protocol to UDP will not generate less traffic,
on busy congested networks it will actually generate more
as it retransmits dropped packets instead of adapting as
TCP does.
Change the mount from "soft" ot "hard" and I bet all your
problems go away...
-David
Vicky Tanya Seno
"nolargefiles" opition. It seem to be making a difference after some time.
So we are not sure if this solved the problem. But if the problem returns,
I'll try going with hard mounting the file systems and maybe try UDP.
Thank You Everyone for your help.
Tanya Seno
Casper H.S. Dik
>Is there anyway to force Solaris to use UDP for NFS mounts. Because at
>the moment most the mount are TCP and its generating to much traffic.
UDP will generate more, not less, traffic
>nfsstat -m
>/home/users1 from nfs1:/home/users1
> Flags:
>vers=3,proto=tcp,sec=sys,soft,intr,link,symlink,rsize=32768,wsize=32768,retrans=5
Don't use softmounts; you might as well take your data out and shoot it.
Casper
--
Expressed in this posting are my opinions. They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.
Benny
>
> Hi All
> Does anyone know what this error means ??
> NFS getattr failed for server nfs1: error 27 (RPC: Received
> disconnect from remote)
>
> We have a new HP machine running HPUX 11i that exports the students
> home directories to all types of sun workstation in the labs. Now all
> the SUN keep getting the above error message.
> Before we had another HP box running 11 export the home directories
> and there were no problems.
>
> Can someone help me out here ? ?
>
> THANK YOU!!
> Tanya
Maybe the NFS server is a bit under stress, do you see these errors all
the time or mainly during daytime when the load is higher ?
If you're using soft mounts then try using hard mounts instead and see
if that helps ?