Miniserver Binding to Wrong Adapter

[griley]

My Sage Server (5.0.4.92 JVM version=1.5.0_10) has 2 network cards. 1 connects to my LAN (192.150.23.200) the other (192.168.0.1) connects *directly* to my new (cool) HDHomeRun encoder.

The 2 MVPs that are on the LAN aren't working because the Sage miniserver has decided to bind itself to to the 192.168.0.1 interface.

By turning on the debugging in the sage.properties file i can see this:

Sat 12/16 0:08:38.257 Trying to find lan network interface
Sat 12/16 0:08:38.268 SageTVDiscoveryServer was instantiated.
Sat 12/16 0:08:38.268 SageTVMiniDiscoveryServer was instantiated.
Sat 12/16 0:08:38.268 SageTVServer was instantiated.
Sat 12/16 0:08:38.262 addr: /127.0.0.1
Sat 12/16 0:08:38.263 addr: /192.150.23.200
Sat 12/16 0:08:38.263 addr: /192.168.0.1
Sat 12/16 0:08:38.263 Miniserver running on 192.168.0.1

and also see that when the MVP tries to connect that Sage is giving the MVP the wrong IP address:

Fri 12/15 23:47:09.765 Received bootp request
Fri 12/15 23:47:09.765 Op:1 hwtype:1 hwaddrlen:6hopcount:0
transaction:-1501536157 nSeconds:100 flags:0
MAC: 00:0d:fe:00:26:f1
Fri 12/15 23:47:09.765 Added MVP client to our bootp list 00:0d:fe:00:26:f1
Fri 12/15 23:47:09.765 servername 192.168.0.1
Fri 12/15 23:47:09.766 Sending bootp reply

I was under the impression that you could tell Sage which IP address the miniserver should bind itself to by setting the proper key in sage.properties

miniserver/gateway=192.150.23.200

but this doesn't seem to affect Sage at all.

If I disable the 192.168.0.1 interface, sage binds to my LAN address and the MVPs boot fine.

Am I missing something? Is Sage 6.x better equipped to handle this issue?

Thanks

[FidgetyRat]

I had a similar problem, except mine was binding to a loopback adapter I added manually.

I'm sure there is some sort of workaround, but since I lack tinkering time I just disabled the annoying adapter. I will lose local recording if my router goes down now, but oh well.

[_Demo_]

You can force it using the property

miniserver/forced_server_ip=1.2.3.4


_Demo_

[griley]

Originally Posted by _Demo_

You can force it using the property

miniserver/forced_server_ip=1.2.3.4


_Demo_

Ya... I discovered this today as a new feature in Sage 6. Another reason to upgrade!

Thanks

[FidgetyRat]

Oh that is a pretty feature. Thanks.

[wvpolekat]

I will be filing a bug report on this also, but Sage basically ignores this setting.

If you use tcpview, you can see that there is no change to what interfaces it listens on when you use this setting.

Sage also has a nasty habit of advertising the wrong IP to clients if there is more than one interface.

In my case, I have 3 network cards and 2 MVPs. One NIC goes to my regular LAN, the other 2 goto seperate networks (and subnets), one for each MVP.

If I leave Sage in the default config, it appears to listen on all interfaces, which is OK.

When an MVP boots, it barks that it is unable to use the previous settings, and scans for streaming servers. Sometimes it finds it, sometimes it doesn't. When it does, the server name is correct but the IP is the one from my home network. If you accept it and go on, it pukes as unable to connect to GUI server.

A packet capture reveals that the MVP broadcasts for the streaming server, so it would not be able to see the IP address that it displays, which tells me that Sage is returning the IP address in the reply. Which, is unreachable to the MVP because it is on a different subnet.

When I use the force setting, it slightly changes the behavior in that it can no longer see anything. I can see it broadcasting looking for the server, but the server never replies. For all intents and purposes, the server is braindead on that network when this setting is enabled.

The only reliable way I have gotten the MVP to work properly 100% of the time is to disable the home LAN interface, which basically makes the Sage server completely inaccessable and unable to get EPG updates and such.

The on conclusion I can draw from this is that both the MVP and Sage are both poorly behaved (at best) in anything other than a plain vanilla network environment. Sage is not smart enough to reply with an address that is on the same subnet as the interface the request came in on. The MVP is not smart enough to look at the address the reply came from for the server address and sanity check that against the one provided by the server or validate that it is on the same subnet.

This is BASIC network 101 stuff and neither of these devices abide by any reasonable standards on how to interoperate.

Sage is a great product, but it's oddities like this that make it not ready for the average joe user.

I would have probably just chocked this up to my overcomplicating things, but the fact that Sage completely ignores a setting makes me mad. If you are going to put a function in, make sure it at least attempts to work, otherwise it serves no purpose but to frustrate people fighting with a red herring.

[stevech]

thanks. really useful sleuthing.

MVP firmware seems to have been poorly designed and Hauppauge is selling enough as is to ignore pleas for a better design.

Wish Sage would support one of the newer/better players. PC World has a review of some in the Aug 07 issue. However the author is fixated on those that support 11g WiFi and HD - as if this makes sense due to bandwidth.

[CollinR]

Is this not a DHCP issue?

Do you have a DHCP server installed to service the MVP clients on that interface?

If not, it's going to find one if it can, once it does it will use that configuration. Basically the MVP will never interface correctly until you setup a HDCP server pointing them to the correct address range.


I also get the other issue in that Sage seems to want to use the same interface for all ethernet traffic. Thats very interesting and may be a bug, you should be able to seperate those networks.

[wvpolekat]

Originally Posted by CollinR

Is this not a DHCP issue?

No, it's not

[QUOTE]
Do you have a DHCP server installed to service the MVP clients on that interface?
[QUOTE]
Yes, there is a seperate scope setup for them and the MVPs as well as the Sage box get IPs from it sucessfully on that subnet.

If not, it's going to find one if it can, once it does it will use that configuration. Basically the MVP will never interface correctly until you setup a HDCP server pointing them to the correct address range.

Correct, but this is not the case.

I also get the other issue in that Sage seems to want to use the same interface for all ethernet traffic. Thats very interesting and may be a bug, you should be able to seperate those networks.

This IS a BIG HUGE SMOKING bug. If the option is to force miniserve to bind to a particular interface, it should do just that. It's clear as day that it does NOT do this. Also, this causes the problem with it returning it's IP that is on a subnet other than the one the request came in on. So, there are actually two bugs.

1) The force miniserve is ignored, it will still bind to all interfaces.
2) Sage does not apply any logic (apparently) in what address it returns to clients. Sage should look at what interface the request came in on and return that interfaces IP to the requesting client. That would eliminate the need for the forced bind in this case. It may still be desired for other reasons, but you could segment traffic without this feature if it provided clients with a sensable IP.

ALL of this can be duplicated at will in just a few min and Sage QC and release management should have caught this.

[Steep]

Originally Posted by wvpolekat

This IS a BIG HUGE SMOKING bug. If the option is to force miniserve to bind to a particular interface, it should do just that.

This post has been edited in order to retard the authors frustration.


<snip>Affirmative</snip>! I can't begin to describe how <snip>aww shucks</snip> I'm getting at the non-stop barrage of <snip>poop</snip> I've had to overcome with my sage setup.

It's not like I'm blindly rushing into uncharted waters, as most of the upgrades and hardware purchases I've made over the last 6 months have been based on recommendations of people here in these forums.

I have the HDHomeRun setup on 1 network card (as many claimed to have working fine), and I have Sage connect to the LAN on another. Ever since, I haven't been able to get my MVP to connect to Sage, but I thought it was something I did wrong on my end until I found this thread.

Why can't <snip>it just work</snip>. Getting so sick of being a lab rat hunting down the <snip>yummy</snip> cheese.

I'm going to submit a bug report on this and <snip>know that it will be fixed in a timely manner</snip>.

[_Demo_]

Could you please give details of your configuration and an ethereal log of the packets that are wrong.

Thanks

_Demo_

[Steep]

Update. Using the "miniserver/forced_server_ip=1.2.3.4"
line in conjunction with stevech's suggestion in this thread (wMVP doesn't play nice with the wireless router in 108mbps mode), seemed to fix the problem. I'm still getting the MVP loading a perma-black screen, but at least it's finding the server now.

[pjwerdna]

For those not on the latest SageTV yet, this thread about Using MS Loopback Driver for UNC path with a HD Homerun (HDHR) might help.