topic Re: Ping issue in Operating System - Linux

Ping issue

WW288996 — Fri, 08 Jan 2010 04:38:22 GMT

Hi,

I have a hp msa storage directly connected to hp dl580 server which has RHEL5.4

I have 2 NIC ports in storage and 6 NIC ports in server.

I have connected directly the storage to server like as follows,

Storage NIC port0 --> Server NIC port eth4
Storage NIC port1 --> Server NIC port eth5

I have assiged IP in same series.

Storage IPs
------------
Storage NIC port0 - 10.0.0.41/255.255.255.0
Storage NIC port0 - 10.0.0.43/255.255.255.0

Server IPs
-----------
Server NIC eth4 - 10.0.0.47/255.255.255.0
Server NIC eth5 - 10.0.0.49/255.255.255.0

My problem is , If I ping from server storage IPs, the first IP is pinging with out mentioning any interface like s follows,
# ping 10.0.0.41
PING 10.0.0.41 (10.0.0.41) 56(84) bytes of data.
64 bytes from 10.0.0.41: icmp_seq=1 ttl=64 time=0.232 ms
64 bytes from 10.0.0.41: icmp_seq=2 ttl=64 time=0.221 ms

My problem is,
But If I ping second storage IP with out mentioning interface its not pingig.
# ping 10.0.0.43
PING 10.0.0.43 (10.0.0.43) 56(84) bytes of data.
From 10.0.0.49 icmp_seq=2 Destination Host Unreachable
From 10.0.0.49 icmp_seq=3 Destination Host Unreachable
From 10.0.0.49 icmp_seq=4 Destination Host Unreachable

But If I specify interface then it will ping like as follows,

# ping 10.0.0.43 -I eth4
PING 10.0.0.43 (10.0.0.43) from 10.0.0.47 eth4: 56(84) bytes of data.
64 bytes from 10.0.0.43: icmp_seq=1 ttl=64 time=0.922 ms
64 bytes from 10.0.0.43: icmp_seq=2 ttl=64 time=0.221 ms
64 bytes from 10.0.0.43: icmp_seq=3 ttl=64 time=0.217 ms

I actually want this ping happens with out specifying interface, because I have to get both ports to configure multipath in RHEL to access the same storage in diffrent paths. As far as I know no gateway is required because it is a direct connection and same series IPs as well.

Pleasae help me in this regard.

Please find the route below, I believe that eth5 comes first thats why it pings with out interface for eth5.

# route
Kernel IP routing table
Destination Gateway Genmask Flags Metric Ref Use Iface
10.0.0.0 * 255.255.255.0 U 0 0 0 eth5
10.0.0.0 * 255.255.255.0 U 0 0 0 eth4
10.58.131.0 * 255.255.255.0 U 0 0 0 eth0
169.254.0.0 * 255.255.0.0 U 0 0 0 eth4
default 10.58.131.1 0.0.0.0 UG 0 0 0 eth0

Re: Ping issue

Steven Schweda — Fri, 08 Jan 2010 05:06:44 GMT

> Ping issue

I'd call it a network configuration problem.

> I have connected directly the storage to
> server like as follows,
>
> Storage NIC port0 --> Server NIC port eth4
> Storage NIC port1 --> Server NIC port eth5

So these are two distinct (physically
separate) networks?

> Storage IPs
> ------------
> Storage NIC port0 - 10.0.0.41/255.255.255.0
> Storage NIC port0 - 10.0.0.43/255.255.255.0
[One of those was really port1?]
>
> Server IPs
> -----------
> Server NIC eth4 - 10.0.0.47/255.255.255.0
> Server NIC eth5 - 10.0.0.49/255.255.255.0

But their IP addresses (and netmasks) say
that they're all on the same subnet
(10.0.0.x)?

That would seem to explain why the system is
confused about where to send the data.

Why not choose a different network address
for each distinct network? For example:

Storage NIC port0 10.0.1.1/255.255.255.0
Server NIC port eth4 10.0.1.2/255.255.255.0

Storage NIC port1 10.0.2.1/255.255.255.0
Server NIC port eth5 10.0.2.2/255.255.255.0

That should give the server a chance of
choosing the right network interface to use
to talk to each port on the storage box.

Of course, many other network+netmask
combinations are possible, but the important
thing is to have different network addresses
for different networks.

Re: Ping issue

WW288996 — Fri, 08 Jan 2010 07:27:20 GMT

To support fault tolerence both should be in the same network.

Re: Ping issue

Steven Schweda — Fri, 08 Jan 2010 09:52:00 GMT

> To support fault tolerence both should be
> in the same network.

Then perhaps you should put them in the same
network.

> I have connected directly the storage to
> server like as follows,
> [...]

That sounds to me like two physical networks.

I don't see how you can have it both ways.

Re: Ping issue

WW288996 — Sun, 10 Jan 2010 15:52:19 GMT

The physical connetion which I have made is direct. It means NIC cables are connected directly from storage to server, its a direct attached storage for my server.

I mean my storage I did not connected to a Switch.

2 cables I used to connect storage ports to server.

Re: Ping issue

Steven Schweda — Sun, 10 Jan 2010 18:13:18 GMT

> The physical connetion which I have made
> is direct. [...]

That's what I thought. Which is why it makes
no sense to me to use the same network
address for these two distinct (physically
separate) networks. (Unless, of course, you
_intend_ to confuse the network software.)

If you tell the system that eth5 can talk to
10.0.0.x, then you shouldn't be amazed when
the system believes you, and tries to use
eth5 to do that.

If you have two different (separate)
networks, then you should probably have two
different network addresses for them.

> To support fault tolerence both should be
> in the same network.

I don't understand that statement, but the
fact remains that you have two different
(separate) networks, not "the same network",
and it makes no sense (to me) to use the one
network address for two different (separate)
networks.