Forums

gumby
gumby
Offline
Resolved
0 votes
Hi all,

I have just switched over from PFSense to ClearOS and have a few questions regarding the Multiwan feature. I have two WAN connections from my ISP and each use DHCP. When I set both up to use DHCP one of the connections always says offline in the Multi-Wan page and when I have a look at /var/log/syswatch it consistantly shows an error setting the gateway

Sun Jun  8 17:53:16 2014  info:    eth1 - restarting DHCP connection
Sun Jun 8 17:53:34 2014 info: eth1 - ping check - no gateway found
Sun Jun 8 17:53:34 2014 info: eth1 - restarting DHCP connection
Sun Jun 8 17:53:50 2014 info: eth1 - ping check - no gateway found
Sun Jun 8 17:53:50 2014 info: eth1 - restarting DHCP connection
Sun Jun 8 17:54:07 2014 info: eth1 - ping check - no gateway found
Sun Jun 8 17:54:07 2014 info: eth1 - restarting DHCP connection
Sun Jun 8 17:54:24 2014 info: system - heartbeat...
Sun Jun 8 17:54:24 2014 info: eth1 - ping check - no gateway found
Sun Jun 8 17:54:24 2014 info: eth1 - restarting DHCP connection


It appears that the solution is to add one of the interfaces as a static connection instead of DHCP. I've never had to do this in PFSense and to my recollection I also never had to do it in ClearOS 5.x. Sometimes I can get both WAN connections working as DHCP but within a few days one of them breaks and goes offline and I need to reset one to static for a little while and then set it back to DHCP and then it magically works again. Settings to static permanently really isn't an option as neither of my IPs given to me by my ISP are static IPs and can/will change over time.

Can anyone explain this behavior? Surely this shouldn't be the expected behaviour.
Monday, June 09 2014, 01:13 AM
Share this post:
Responses (35)
  • Accepted Answer

    Friday, January 16 2015, 03:10 AM - #Permalink
    Resolved
    0 votes
    Well it is up and running. And working, otherwise, I couldn't be sending this as that D525MF system is now the "production" gateway (production level test)
    The reply is currently minimized Show
  • Accepted Answer

    Wednesday, January 14 2015, 08:16 PM - #Permalink
    Resolved
    0 votes
    Ok, re-installed from scratch. Then immediately updated with the correct rpm for the nics and rebooted.

    Then installed all the software from the marketplace that I wanted, except for multi-wan.

    Got it up and working, having removed a Trendnet "smart" switch (old thing I should have used a sledge hammer on...).

    System works. So I changed it to use 8.8.8.8 & 8.8.4.4 for DNS.

    Ok, I did some hardware swapping and got my Netgear 8 port smart switch between this gateway and the cable modem.

    At this point, eth0 is WAN and eth1 is LAN, eth2/3 are not connected.

    So, I added eth3 as EXT getting its address via DHCP and this NIC to use a different hostname and to NOT autoupdate DNS.

    Eth3 makes connections and we see that Gateway, DNS, & Internet all connect. In a few moments it gets a spinning wheel. And then that resolves.

    Meanwhile, on an attached computer I have been doing pings and tracerts to see what is happening.

    Now and then they will fail and then when retried they work.

    So I pulled the cable from eth0 and then told multi-wan to delete it.

    It had already switched to eth3, but would get the spinning wheel now and then. So I deleted eth0 and changed the host name of eth3. After that, the spinning wheel has ceased.

    So far so good.

    Next test is to set up the DHCP to have the same configuration the one I use for the whole network, and swap it in and see if everything works from there.
    The reply is currently minimized Show
  • Accepted Answer

    Wednesday, January 14 2015, 08:08 AM - #Permalink
    Resolved
    0 votes
    Just about all distros have the wrong drivers as the r8168 driver is not built into the kernel and the r8169 driver, which is in the kernel, is flagged as compatible with RTL8111/8168 devices.
    The reply is currently minimized Show
  • Accepted Answer

    Wednesday, January 14 2015, 04:05 AM - #Permalink
    Resolved
    0 votes
    Well, things are so twisted? that I decided to re-install ClearOS from scratch. Now if it can see eth0 connected to the cable modem here, why can't it run with it? I'm starting to suspect the multi-wan code and this MOBO do not play well together. Until tomorrow...
    The reply is currently minimized Show
  • Accepted Answer

    Wednesday, January 14 2015, 03:10 AM - #Permalink
    Resolved
    0 votes
    I had spec'ed out a motherboard with a US based distributor. They said here use this Gigabyte board (wow did I suffer sticker shock when the told me how much). Then were not happy with having to take that board back. I told them about the number of complaints I had found on the internet about it. And how ludicrous Gigabyte's support was. They seriously told me that the BIOS had the clarvoiyant option -- OK, not in those words, but you can't tell what is in a DVD until you read from it. Ok, I'm beating a dead horse here.

    The D525MF was working. Now it is not. In the system log I can see that it is looping. It gets to the point where it wants to do NTP sync. This fails with "Error resolving time.clearsdn.com: Name or service not known.

    This is starting to look like a DNS problem.

    So I will change to running off my working gateway and see if it can resolve from there.
    The reply is currently minimized Show
  • Accepted Answer

    Wednesday, January 14 2015, 02:08 AM - #Permalink
    Resolved
    0 votes
    Gigabyte's GA-J1900N-D3V again :-( see http://www.clearfoundation.com/component/option,com_kunena/Itemid,232/catid,24/func,view/id,62006/#62006 for a previous failure in these forums... Gigabyte really messed up the BIOS on this one... do a search on the internet. Many frustrated buyers and a precious few who managed to get it running something other than windows... including the contortions some went though to be able to either update the BIOS or change settings.

    The D525MF looks more interesting, though the information on the web other than from suppliers - Alibaba :-) is almost totally non existent. Looking at the TEKET web-site motherboard section it is interesting to note that scattered throughout the list are also Gigabyte, ASUS, Intel and VIA original boards including the GA-J1900N-D3V. This makes me wonder whether the TEKET brand name boards are either built by one or more of these companies, or designed by one or more of them and manufactured by TEKET under some sort of agreement... Their support page is dismal - just a list of worldwide contacts. No downloads for manuals, drivers, BIOS updates and BIOS update software, etc - absolutely nothing.

    Good Luck with the D525MF...
    The reply is currently minimized Show
  • Accepted Answer

    Tuesday, January 13 2015, 10:48 PM - #Permalink
    Resolved
    0 votes
    Tony Ellis wrote:
    Can you name the windows-only product and also the motherboard you are using with 4 NICs for testing so other users here are alerted to potential problems?.


    Gigabyte's GA-J1900N-D3V
    TEKET Innovation Technology Co. LTD's D525MF -- In fairness, ClearOS had the wrong drivers for the RTL NICs. I have corrected that.

    Both of these companies tell you that you can run Linux on their systems and that they have drivers, etc. But that Gigagbyte board was just too funny. At least I could send it back with an RMA. International orders are a bit more involved.

    But then, that is why we order samples to do testing before committing. In this case, we are probably going to drop this whole thing. I have 4 months in this project, expecting to go "GA" [Generally Available] in March 2015. It ain't gonna happen. I just do not have the time to spend since i work full time on z/frames (IBM Mainframes) doing capacity planning and tuning for a large insurance company.


    Ok, I will be back later this evening to work on the D525MF board and see if I can get it to behave better.
    The reply is currently minimized Show
  • Accepted Answer

    Tuesday, January 13 2015, 03:31 PM - #Permalink
    Resolved
    0 votes
    Interesting... I will never allow myself to be a guinea big with hardware :-) only software that I can delete if proved useless... and always research hardware first to make sure it works with the intended OS, and if a mother-board that all the options on the motherboard work (unless it is something not required and can be disabled in BIOS).

    Can you name the windows-only product and also the motherboard you are using with 4 NICs for testing so other users here are alerted to potential problems?

    Also I would suggest you are using the term "hub" when in fact you mean switch. No modem to my knowledge today that provides NAT to 2 or more LAN connections would have a hub. They all have built-in switches... See http://www.ccontrols.com/enews/0211story2.htm - one of many hits that will explain the difference. With massive quantities and better higher density silicon the cost of switches is no longer an issue. The only place for a hub is for network diagnosis.

    As for two paths leading eventually to the same modem, they eventually come together on the external interface. I wonder if the return packets are not always routed correctly once the internal split is encountered. Just a thought... I know that in certain circumstances (https?) multi-wan can stumble when an external site finds packets from the same system arriving via different external addresses. I have also experienced this very occasionally with some downloads failing to start. This is resolved by a script I have that temporarily gives one interface maximum weight and the other minimum. If you use multi-wan fail-over rather than balance this will not be a problem.

    As for the 192.168.100.1 "problem" you simply create two disabled static "via" routes - one for each external interface and simply enable the one you want. More crudely, since this would probably be a rare use case, unless you run a routine job to collect statistics, just turn off the modem that you don't want to connect to.

    As for sub-nets to use, Nick's advise is wise indeed. You need to plan your network talking into consideration what devices and network software you intend to use. In my case 182.168.0.0 and 192.168.4.0 are external
    and 192.168.2.0, 192.168.1.0, 192.168.3.0 and 192.168.5.0 (all /24) for internal all work perfectly for me - but obviously not for every-one...
    The reply is currently minimized Show
  • Accepted Answer

    Tuesday, January 13 2015, 02:22 PM - #Permalink
    Resolved
    0 votes
    Tony Ellis wrote:
    What is all this testing supposed to achieve? Your testing set-up in my mind is too far removed from a real multi-wan that it is not worth the trouble, and will not alert you to other potential problems. A case in point is DNS. See Nick's append on this http://www.clearfoundation.com/component/option,com_kunena/Itemid,232/catid,19/func,view/id,65242/#65243. Your test, if you should get it to work, would not show up this problem as both WANs would end up at the same ISP and same DNS servers.


    [I had a prospective customer who wanted the multi-wan support. But due to things beyond my control, we have blown the deadlines and lost the account.]

    I am testing different motherboards that have 4 or 6 NICs built in. Having worked in the hardware world for a computer vendor or three, I know that there are situations where something is not right with an off the shelf mobo. And you can't get it to work.

    Example: I recently had a MOBO that would only run W/7 or W/8!! And that was IF and ONLY IF you could get it to come out of some pre-BIOS shell to actually do POST. We sent that mobo back as being defective and the manufacturer was screeming about us not using a Windows install. :huh: [They are apparently using Clarvoyant BIOS: it knows that the DVD in the USB drive isn't a Windows DVD.] ;)

    This particular MOBO I'm testing (4 NICs) has exhibitted some interesting issues. It worked for about 30 minutes and then started this "renegotiation" thing in a rolling process for any NIC set to EXT. Updated the drivers, and it still does it.

    If I only have 1 NIC attached to EXT it works fine.

    So if I have eth0 connected to the cable modem, and eth3 connected to my firewall gateway, it still does not work (completely different subnets) -- BTW if I remember correctly, this was the configuration that worked for about 30 minutes before I changed configurations to test across all 4 NICs to ensure there is not some special configuration that will NOT work.

    BTW -- I am forcing DNS to use 8.8.8.8 / 8.8.4.4 so that it doesn't make any difference which WAN is being used.

    Nick Howitt mentioned cable modems operating in Bridge mode.

    My ISP is Bright House (a RR user). The cable modems they provide are also "hubs". So the moment you plug into other than eth0, it goes into NAT mode.

    The [A]DSL modem I've recently used provides CLASS C PVT w/ DHCP. The AT&T UVerse with 2WIRE "cable modem" that I've worked with is a firewall, WiFi and 4 port hub(?).

    All of these have been provided to businesses. And I would suggest to a Business that they obtain their own DOCSIS 3 compliant Cable modem so that they can configure it as they need to.
    The reply is currently minimized Show
  • Accepted Answer

    Tuesday, January 13 2015, 12:36 PM - #Permalink
    Resolved
    0 votes
    In my minimal experience with one ISP, pure cable modems have always worked in bridge mode so have passed on the WAN IP to ClearOS rather than operated in NAT mode. The problem with them and MultiWAN is that their LAN connected address has always been 192.168.100.1 and is not user configurable so if you try to connect to one, for example to read its configuration, it would be random which one you connected to.

    On the other hand when the ISP has supplied a combined modem/router (the Virginmedia SuperHub which is a Netgear device under the skin) it has come preconfigured with a LAN IP and subnet (from memory 192.168.0.1/24 for LAN .1.0/24 for isolated wireless and .2.1/24 for guest wireless), but this is user configurable. This means that there would be no problem with two of these devices connected to different cables as I can freely change the LAN IP and subnet. If I put the modem/router into bridge mode it then changes its internal IP to 192.168.100.1 which is not user configurable.

    One thing I'd like to point out is that, if you can, you should always try not to use 192.168.0.0/24, 192.168.1.0/24, 192.168.100.0/24 (and now with the VM SuperHub, 192.168.2.0/24) anywhere in your network. They are relatively common and will potentially mess up any future use of roadwarrior VPN's. Then can also lead to other issues if you add more networking kit to your network.
    The reply is currently minimized Show
  • Accepted Answer

    Tuesday, January 13 2015, 06:25 AM - #Permalink
    Resolved
    0 votes
    Nick is correct - you cannot run multi-wan with each WAN interface being in the same sub-net and having the same gateway address - just impossible...

    I have two ISPs. One modem is sub-net 192.168.0.0/24 gateway 192.168.0.1 - the other sub-net 192.168.4.0/24 gateway 192.168.4.1 - works perfectly.


    [root@may ~]# ip route
    192.168.4.0/24 dev eth0 proto kernel scope link src 192.168.4.34
    192.168.2.0/24 dev eth1 proto kernel scope link src 192.168.2.34
    192.168.0.0/24 dev eth2 proto kernel scope link src 192.168.0.34
    default via 192.168.0.1 dev eth2
    [root@may ~]#

    192.167.2.0/24 is my LAN sub-net.

    What is all this testing supposed to achieve? Your testing set-up in my mind is too far removed from a real multi-wan that it is not worth the trouble, and will not alert you to other potential problems. A case in point is DNS. See Nick's append on this http://www.clearfoundation.com/component/option,com_kunena/Itemid,232/catid,19/func,view/id,65242/#65243. Your test, if you should get it to work, would not show up this problem as both WANs would end up at the same ISP and same DNS servers.

    see this rather long thread http://www.clearfoundation.com/component/option,com_kunena/Itemid,232/catid,19/func,view/id,55111/ for earlier problems with multi-wan and in the later section changes I suggested that solved problems in certain situations for the version of the software then current...

    Another thing to remember with multi-wan is that routes are cached, so if a lot of your traffic goes to the same destination, it will tend to always favour the same WAN.
    The reply is currently minimized Show
  • Accepted Answer

    Tuesday, January 13 2015, 03:31 AM - #Permalink
    Resolved
    0 votes
    But this is a valid multi-wan situation. Eth0 is being handled by ABC company and eth3 is being handled by xyz company.

    Since each cable modem is being set for NAT, and then has to do DHCP, it also must also send all unknown addresses to the cable modem (Gateway), and so if both cable modems are set up the same, you will get 192.168.0.1 as the gateway and DHCP.

    In a Multi-Wan situation, this should work. But since it appears to not work, how would one fix this?

    The only way I know would be to have the ISP set their cable modem to bridge. Then you should be getting CLASS A public or a CLASS B private -- but you should not be getting CLASS C private with 192.168.0.* on both cable modems (or [A]DSL modems) in that case.

    So in order to circumvent the situation where both NICs have the same gateway address (193.168.0.1), I set the one NIC to connect to my actual firewall/Gateway (ClearOS) as its ISP. And the other I connected direct to the cable modem. In this configuration you get one NIC at 192.168.1.? while the other is at 192.168.0.? .
    Multi-Wan still doesn't work.

    And so this last is the configuration where I will start tomorrow evening when I get everything all set up again.
    The reply is currently minimized Show
  • Accepted Answer

    Monday, January 12 2015, 10:20 PM - #Permalink
    Resolved
    0 votes
    I don't think it is valid to have two identical gateway addresses which are the same gateway on different NIC's. That is a routing loop. Both NIC's should be on different subnets or ClearOS does not have a chance of working out which NIC to use to send the traffic and many ethernet switches will disconnect under those conditions.
    The reply is currently minimized Show
  • Accepted Answer

    Monday, January 12 2015, 07:56 PM - #Permalink
    Resolved
    0 votes
    I will set that unit back up for testing, probably tomorrow.

    Meanwhile, suppose that ISP A and ISP B happen to use the same PVT CLASS C subnet?

    Let us say that eth0 is WAN(A) and eth3 is WAN( B ).

    So eth0 is looking at gateway address of 192.168.0.1 and gets the address of 192.168.0.10.

    Then eth3 is looking at a gateway address of 192.168.0.1 and gets the address of 192.168.0.5.

    Will we get this same problem?

    I had thought that this could be a valid situation. So I set up to have two "taps" off my cable modem to test the behavior. Each ping that is issued should be routed to the correct address (192.168.0.1) with the IP address of the NIC from where the ping is sent.

    If this is not a valid situation, then the only way to prevent this is to set the cable modem(s) to BRIDGE mode, not NAT.
    The reply is currently minimized Show
  • Accepted Answer

    Monday, January 12 2015, 12:41 PM - #Permalink
    Resolved
    0 votes
    Can you have a look at /var/log/syswatch and see what it is reporting? I think you can set the ping servers manually in /etc/syswatch.conf or something like that. It may be worth trying that as well, making sure that the three servers are all reachable from both WAN's.

    I am concerned about two things. Firstly I am not sure if you have a routing loop or not having effectively two parallel paths to the internet. Part of me thinkg thisis wrong and another part of me says this is fine. Secondly, in terms of syswatch, the first ping server is normally your WAN's gateway IP. The WAN's gateway IP for one NIC will not be available to the other NIC.

    What is the output of
    ifconfig | grep eth -A 1
    cat /etc/clearos/network.conf
    The reply is currently minimized Show
  • Accepted Answer

    Sunday, January 11 2015, 10:36 PM - #Permalink
    Resolved
    0 votes
    Ok, the rpm install says it completed successfully.

    So I rebooted and did connnections and now it is recycling eth0 then eth3 then eth0... (those are the two set for EXT). Eth1 and eth2 are LAN and eth1 is 192.168.5.1 while eth2 is 192.168.6.1.

    lsmod shows r8168 was loaded with the kernel.

    I've also changed eth0 to go to my gateway before getting to the cable modem. Still no joy. Eth0/3 keep recycling.

    Anyone have any other ideas?
    The reply is currently minimized Show
  • Accepted Answer

    Sunday, January 11 2015, 09:36 PM - #Permalink
    Resolved
    0 votes
    Sorry for delay. Finally got back to this. Got it mounted and now going after rpm install....
    The reply is currently minimized Show
  • Accepted Answer

    Tuesday, January 06 2015, 12:04 PM - #Permalink
    Resolved
    0 votes
    Look at the output of dmesg to find out what it is called. I would have thought that mounting by label should work but otherwise use the device name from dmesg.
    The reply is currently minimized Show
  • Accepted Answer

    Tuesday, January 06 2015, 02:45 AM - #Permalink
    Resolved
    0 votes
    Thanks to weather here, I was able to get back to this tonight.

    I loaded the rpm on a thumb drive. I can't seem to get it to mount on the test system.

    Is there something I need to know about ClearOS and USB drives? I can't get it to mount by Label, and I can't seem to deduce what "drive" it is.
    The reply is currently minimized Show
  • Accepted Answer

    Monday, January 05 2015, 07:23 PM - #Permalink
    Resolved
    0 votes
    Steve Thompson wrote:
    Yes. I was trying to run off eth2, or eth1, at different times, as the LAN NIC.
    But your log indicates problems with eth2. I am just trying to eliminate it from the equation be defining it with a fixed IP in a different subnet from eth1.

    [edit]
    ... and my later r8168 driver can be found here
    [/edit]
    The reply is currently minimized Show
  • Accepted Answer

    Monday, January 05 2015, 02:06 PM - #Permalink
    Resolved
    0 votes
    Yes. I was trying to run off eth2, or eth1, at different times, as the LAN NIC.

    I will have to come back to this later -- I can't work on it until tomorrow evening because of prior commitments (and besides, my primary job is on z/frames...).

    Thanks for the link. I've captured it to my Suse workstation -- I work from home, so I do get to look at my email now and then :-) .
    The reply is currently minimized Show
  • Accepted Answer

    Monday, January 05 2015, 12:16 PM - #Permalink
    Resolved
    0 votes
    There is a pre-compiled driver here. Install it and reboot. I have a slightly later verson compiled, but I can't search the link at work as it is blocked.

    Have you tried explicitly setting eth2 to LAN?
    The reply is currently minimized Show
  • Accepted Answer

    Monday, January 05 2015, 12:21 AM - #Permalink
    Resolved
    0 votes
    I expect, for the WAN, using the cable modem, 192.168.0.?. When using the Gateway, 192.1681.?

    This test system is to use, for LAN 192.168.5.1 and hand out 192.168.5.? for IP addresses.

    It is a 64bit ClearOS installation. And I was not sent drivers with this MOBO, or any doc for it (that's another story). Vendor apparently took 23DEC to 05JAN off.

    I'll go find the drivers... I hope they are set up to "self make". I do not have this system set up for development. It is a test system to see if their boards work for us.
    The reply is currently minimized Show
  • Accepted Answer

    Sunday, January 04 2015, 09:45 PM - #Permalink
    Resolved
    0 votes
    What IP addresses are you expecting at your WAN NIC's and on which subnets? Also what is the LAN subnet for Eth1?

    As a secondary issue, did you install the r8168 drivers for your NIC's? By default they use the r8169 driver which is not good. If you did not, are you running the 32bit or 64bit version of ClearOS (...or what is the output of "uname -r")?
    The reply is currently minimized Show
  • Accepted Answer

    Sunday, January 04 2015, 09:09 PM - #Permalink
    Resolved
    0 votes
    Ok lemme answer all the questions at once.

    Why do I want to do this? Well, I generally try to test things before hooking them up as the "production" system. So, as best I can I am trying to simulate a multiple ISP situation.

    Here is my configuration: Cable Modem fed by RoadRunner. This feeds to my "production" gateway, which is a 2 nic system running ClearOS. This then feeds my whole LAN and connects to my WiFi equipment via a smart switch.

    Now the test system that is the subject of discussion, is a 4 NIC MOBO system (4 RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller). I currently have eth0 connected to the production gateway via two smart switches (one at my desk, the other over at the production gateway). Eth3 is connected to the cable modem via a 4 port Trendnet hub.

    Multi-Wan shows that eth3 is offline. It will not obtain online status, while eth0 shows it is online. [more to this later]

    Question about recycling the Cable Modem. Unless you know something I don't, this modem is in NAT mode and will handle multiple connections (already tried and tested prior to this point, in testing a different gateway router...).

    Question about how connections are noted/configured: When I connect something as an EXTernal connection, I so note it to ClearOS in the IP Settings, Network Interfaces area. I also make sure, to avoid "double natting" to make the in use LAN connection (i'm only using one, the other is disabled) to use a different subnet.

    ------
    Update based on continued experimentation:

    I now find it interesting that where I had this working, one EXT nic going to the cable modem, one EXT going through my production gateway, it is now NOT working. First one EXT then the other EXT show an attempt to reconnect/restart connections. This is the same thing I saw when I had both EXT nics connected to the Cable Modem.

    So I tried setting the hostname for the EXT Nics to be different in the case that this was a problem. This did not fix the issue, so that is not the problem

    I have also noticed other postings discussing this same thing (can't get both links running).

    >>>> This is preventing me from connecting this to my LAN so that I can capture screen shots to attach. OR to get to the internet so I can email captured screen shots from my laptop to my Suse desktop system where I am writing this posting. <<<<

    If I mark eth3 disabled/offline, then I get connections that "work" according to the status (Gateway, Internet, DNS) and then they all go red.

    Looking at the log, I notice that there is an interesting line:
    if_get_address: SIOCGIFADDRE: Cannot assign requested address
    Failed to detect IP Address for WAN interface eth3
    WARNING: No configured WAN interfaces, continuing anyway...
    Warning LAN interface is not configured: eth2
    Execution time: 0.008s
    Synchronizing multipath routing tables...
    Loading environment
    Assuming device is a LAN interface: eth2
    Detected WAN role for interface: eth3
    Detected LAN role for interface: eth1
    Detected LAN role for interface: eth2
    Warning: LAN interface is not configured: eth2

    So what would cause this inability to assign a requested address? (would this have been before I could get in and change the LAN subnet from ..1.1 to ...5.1?

    After this there is a loop, and it basically says that there is no WAN interfaces up or configured, not starting bandwidth manager, attempts to start things and comes back to this message and does it over and over.

    Rebooting does not seem to get it to resynchronize.

    Any ideas?
    The reply is currently minimized Show
  • Accepted Answer

    Sunday, January 04 2015, 06:22 PM - #Permalink
    Resolved
    0 votes
    Oops. I did not realise it was not Steve Thompson who had replied.
    The reply is currently minimized Show
  • Accepted Answer

    Sunday, January 04 2015, 06:08 PM - #Permalink
    Resolved
    0 votes
    Nick - my post was in reply to the original poster's1st two posts in this thread where it sounded like he was swapping connected NICs to the modem and only one was getting DHCP response which could happen if the modem wasn't cycled.

    Peter
    The reply is currently minimized Show
  • Accepted Answer

    Sunday, January 04 2015, 06:02 PM - #Permalink
    Resolved
    0 votes
    I am aware of the CM MAC thing and I am not sure how it applies to what I said. I was just suggesting that if you have any NIC not connected to anything, it should not be defined as WAN.

    What is your topology? Do you have more than one WAN NIC connected to your modem. If so, why?
    The reply is currently minimized Show
  • Accepted Answer

    Sunday, January 04 2015, 05:30 PM - #Permalink
    Resolved
    0 votes
    Not sure if applicable in this scenario but, for many cable providers, anytime the connected MAC address changes (listed as CPE - customer provided equipment - in the cable modem setup) the cable modem needs to be rebooted or it will need answer the DHCP request.

    Peter
    The reply is currently minimized Show
  • Accepted Answer

    Sunday, January 04 2015, 07:21 AM - #Permalink
    Resolved
    0 votes
    Do you have more than one NIC defined as external? If so are they all connected externally? If there are any unconnected NIC's can you try defining them as LAN NICs on different subnets rather than external?
    The reply is currently minimized Show
  • Accepted Answer

    Sunday, January 04 2015, 02:12 AM - #Permalink
    Resolved
    0 votes
    I have a similar problem. And I'm glad to see someone else having a related issue using adapter cards.

    I have a MOBO with 4NICs built in (actually built into the MOBO). I have another with 6NICs, but I'm not able to test with it right now.

    In my case, with the 4NIC mobo, if I have eth0 and ethx (pick 1 of nics 1-3) connected to the cable modem that I have, they will both cycle -- apparently renegotiating their connections. I have also tested with eth2 and eth3.

    Before I had tried this, I had the current test system connected to my working system on eth0, and eth3 connected to the cable modem. In that configuration this problem did not occur.

    My question is, could it be that with both adapters being defined as stonewall.ccccccnet if this is causing the cable modem to have heartburn or is it causing ClearOS 65.0 (final) to have problems?

    I can't get to the cable modem to see if this is its problem (need special logon?). But it is interesting that if eth0 goes through another gateway to the cable modem and eth3 goes direct (or vice versa), then I can simulate a two ISP connection.

    So, how can I tell if this is a ClearOS issue, cable modem issue, or needing to re-id one of the NICs?
    The reply is currently minimized Show
  • Accepted Answer

    gumby
    gumby
    Offline
    Tuesday, June 10 2014, 04:42 PM - #Permalink
    Resolved
    0 votes
    Thanks Peter.

    Let me know if I can provide the same output for when both WAN connections are using DHCP and the one connection is offline (constantly renewing because the gateway can not be found).
    The reply is currently minimized Show
  • Accepted Answer

    Tuesday, June 10 2014, 03:41 PM - #Permalink
    Resolved
    0 votes
    Thanks for the output. We'll be reviewing this in Wednesday's tech meeting. More to come.
    The reply is currently minimized Show
  • Accepted Answer

    gumby
    gumby
    Offline
    Tuesday, June 10 2014, 03:07 AM - #Permalink
    Resolved
    0 votes
    Hi Peter,

    Here is the output. This is when I have eth0 as static and eth1 and dhcp.

    [root@gateway ipfm]# ip route
    192.168.0.0/24 dev eth2 proto kernel scope link src 192.168.0.253
    11.22.33.0/22 dev eth1 proto kernel scope link src 11.22.33.133
    22.33.44.0/22 dev eth0 proto kernel scope link src 22.33.44.96
    default via 11.22.33.1 dev eth1


    eth0 will only work with a static IP for so long and only after I aquire the IP via DHCP first from my ISP (After which I set manually in ClearOS - IP Setting to avoid the issue with eth0 going offline). Sometimes both eth0 and eth1 will remain online using DHCP only for one to fail within a day or so (lease time perhaps?). From memory, it has been the device to renew its lease from the ISP first but I have not documented this.
    The reply is currently minimized Show
  • Accepted Answer

    Tuesday, June 10 2014, 01:06 AM - #Permalink
    Resolved
    0 votes
    Hi gumby.

    Could you post your routing table ("ip route" command)? Feel free to obfuscate the WAN IP of course.
    The reply is currently minimized Show
Your Reply