Hello,
I have a machine running Debian Linux as router and NFS server, with two network interfaces, one connected to a big switched LAN (called eth1) and the other to an isolated LAN (called eth0). The problem is that since the day one interface (eth1) changed it's IP address (due to DHCP) the NFS clients in the isolated LAN (therefore accesing via eth0 to the server) began to have problems mounting the exported directories. The NFS clients sometimes succeed in mounting the directory after a long delay and sometimes they fail due to a RPC timeout. I have captured the network traffic and it seems to be the server which is not responding to the mount call (RPC v.2 MNT v.1 CALL).
There are several strange things I've tested:
- When I take down eth1 (the interface which changed IP address) the clients can mount the NFS server perfectly without any delay. However if I unplug the network cable without doing an "ifdown", it doesn't work. What's more, if I force the old IP address which eth1 had, it works perfectly (!!).
- I've changed the NICs from two Intel 82540EM Gigabit Ethernet to two 3Com 3c905B (and therefore the kernel modules) but the behaviour remains the same.
- I've tested it with other kernel versions (from 2.6.11 to 2.4.25 and a default 2.6.11) and it remains the same.
- I've stopped all non-nfs-related daemons and it remains the same.
- I've looked for that old IP address inside all configuration files, but it doesn't appear.
- Finally I've tested in another machine, the same model as the other server, but running a RedHat FC4 with it's NFS server, and neither it works!!.
Here'sI a software and hardware description:
- Machine: Dell Optiplex GX 270 with two Intel network interfaces (both Intel 82540EM Gigabit Ethernet), one PCI and the other integrated in the motherboard.
- O.S.: Debian Testing with Linux 2.6.11 kernel, with nfs-common package (from apt) version 1:1.0.7-3 and nfs-kernel server package (from apt) version 1:1.0.7-3.
I can't understand what's happening!. Does anyone know where's the problem?.
Thank you in advance,
Rafael.
The only thing that comes to mind is check the dns setup ie
(/etc/resolv.conf) and
see what it is doing.
If dns is slow lots of things get rather messy, and also make sure that the
addresses
you are using on the internal network is a proper public network, and not
one that
is owned by someone else, as if it is someone elses you machine will attempt
to hit their dns server every time.
Roger
Atipa Technologies
_____
From: [email protected]
[mailto:[email protected]] On Behalf Of Barriuso Maicas,
Rafael
Sent: Thursday, June 23, 2005 10:00 AM
To: [email protected]
Subject: [NFS] Multiple NICs server and NFS mount problem
Hello,
I have a machine running Debian Linux as router and NFS server, with two
network interfaces, one connected to a big switched LAN (called eth1) and
the other to an isolated LAN (called eth0). The problem is that since the
day one interface (eth1) changed it's IP address (due to DHCP) the NFS
clients in the isolated LAN (therefore accesing via eth0 to the server)
began to have problems mounting the exported directories. The NFS clients
sometimes succeed in mounting the directory after a long delay and sometimes
they fail due to a RPC timeout. I have captured the network traffic and it
seems to be the server which is not responding to the mount call (RPC v.2
MNT v.1 CALL).
There are several strange things I've tested:
- When I take down eth1 (the interface which changed IP address) the clients
can mount the NFS server perfectly without any delay. However if I unplug
the network cable without doing an "ifdown", it doesn't work. What's more,
if I force the old IP address which eth1 had, it works perfectly (!!).
- I've changed the NICs from two Intel 82540EM Gigabit Ethernet to two 3Com
3c905B (and therefore the kernel modules) but the behaviour remains the
same.
- I've tested it with other kernel versions (from 2.6.11 to 2.4.25 and a
default 2.6.11) and it remains the same.
- I've stopped all non-nfs-related daemons and it remains the same.
- I've looked for that old IP address inside all configuration files, but it
doesn't appear.
- Finally I've tested in another machine, the same model as the other
server, but running a RedHat FC4 with it's NFS server, and neither it
works!!.
Here'sI a software and hardware description:
- Machine: Dell Optiplex GX 270 with two Intel network interfaces (both
Intel 82540EM Gigabit Ethernet), one PCI and the other integrated in the
motherboard.
- O.S.: Debian Testing with Linux 2.6.11 kernel, with nfs-common package
(from apt) version 1:1.0.7-3 and nfs-kernel server package (from apt)
version 1:1.0.7-3.
I can't understand what's happening!. Does anyone know where's the problem?.
Thank you in advance,
Rafael.
Thank you very much for the hint, but I've tested it emptying the resolv.conf file and it still doesn't work.
Rafael.
-----Mensaje original-----
De: Roger Heflin [mailto:[email protected]]
Enviado el: jueves, 23 de junio de 2005 17:19
Para: Barriuso Maicas, Rafael; [email protected]
Asunto: RE: [NFS] Multiple NICs server and NFS mount problem
The only thing that comes to mind is check the dns setup ie (/etc/resolv.conf) and
see what it is doing.
If dns is slow lots of things get rather messy, and also make sure that the addresses
you are using on the internal network is a proper public network, and not one that
is owned by someone else, as if it is someone elses you machine will attempt
to hit their dns server every time.
Roger
Atipa Technologies
_____
From: [email protected] [mailto:[email protected]] On Behalf Of Barriuso Maicas, Rafael
Sent: Thursday, June 23, 2005 10:00 AM
To: [email protected]
Subject: [NFS] Multiple NICs server and NFS mount problem
Hello,
I have a machine running Debian Linux as router and NFS server, with two network interfaces, one connected to a big switched LAN (called eth1) and the other to an isolated LAN (called eth0). The problem is that since the day one interface (eth1) changed it's IP address (due to DHCP) the NFS clients in the isolated LAN (therefore accesing via eth0 to the server) began to have problems mounting the exported directories. The NFS clients sometimes succeed in mounting the directory after a long delay and sometimes they fail due to a RPC timeout. I have captured the network traffic and it seems to be the server which is not responding to the mount call (RPC v.2 MNT v.1 CALL).
There are several strange things I've tested:
- When I take down eth1 (the interface which changed IP address) the clients can mount the NFS server perfectly without any delay. However if I unplug the network cable without doing an "ifdown", it doesn't work. What's more, if I force the old IP address which eth1 had, it works perfectly (!!).
- I've changed the NICs from two Intel 82540EM Gigabit Ethernet to two 3Com 3c905B (and therefore the kernel modules) but the behaviour remains the same.
- I've tested it with other kernel versions (from 2.6.11 to 2.4.25 and a default 2.6.11) and it remains the same.
- I've stopped all non-nfs-related daemons and it remains the same.
- I've looked for that old IP address inside all configuration files, but it doesn't appear.
- Finally I've tested in another machine, the same model as the other server, but running a RedHat FC4 with it's NFS server, and neither it works!!.
Here'sI a software and hardware description:
- Machine: Dell Optiplex GX 270 with two Intel network interfaces (both Intel 82540EM Gigabit Ethernet), one PCI and the other integrated in the motherboard.
- O.S.: Debian Testing with Linux 2.6.11 kernel, with nfs-common package (from apt) version 1:1.0.7-3 and nfs-kernel server package (from apt) version 1:1.0.7-3.
I can't understand what's happening!. Does anyone know where's the problem?.
Thank you in advance,
Rafael.