Hi everyone,
I was wondering if any of you have experience with cable modem 3.0 that suddenly lost Internet connection for sometime, like 30 minutes, one hour or more. I am having that issue with some random clients and when we check their parameters everything is normal, but the clients keep saying the same: they lost connection, even though all the lights on the CM are normal.
Could it be related to load balancing or any misconfiguration on the CMTS? We have an UBR10012, but this is the first time facing this kind of issue.
I'm the last in the group as far as the 10k goes. but something I would look into is arp and lease time. increase the arp timeout, dhcp for modems should be a few weeks instead of hours for example.
when the modem stops passing data, case of you can't ping the modem, but can do a docsis ping?
number of modems on a given downstream/upstream increased and utilization is too high?
snr drops too low or burst noise issues?
Thanks for taking time to help me with the issue. I increase the lease time from 10 minutes to 21 days and that has helped me a lot with some issues of many clients. Didn't know a low lease time could cause so much trouble, but there are still a few clients that are experiment drop connection randomly for few minutes or seconds. Some of the clients, for example, are watching instagram stories and suddenly the receive an error message: "We are having trouble loading this live video" or if they are connected to an online class it suddenly take them out of the session.
I haven't test the pings, in the clients where I am having more issues the bonding is 4x2 and the upstream SNR in on channel is 25.3 dB and in the other 32.6 dB, the number of modems is few like 19 modems on that bonding.
If the lease time was 10 minutes, perhaps the problem isn't the cmts, but you are overloading your dhcp server! We at one time were using Cisco CNR version 3.5, and I believe it could only give out about 40 leases per minute. Graduating to version 6.5 changed things drastically, however there are still buffers for modems waiting in queue to refresh/renew their leases.
I am unfamiliar with any other versions of dhcp servers, but I know what hardware that it is running on is quite important as well!
cable modems are on a private network, example 10.10.10.x there's no need for minutes, rather weeks is recommended. remember, a lease of 2 minutes will send a request to keep the ip to the dhcp server every minute, half the lease time it's set for. like Cableguy said, your server at one point won't be able to keep up.
snr of 21 and higher is needed for 16Qam, snr of 25 for 32Qam and snr of 28 for 64Qam (minimum).
monitor your snr and make sure you have the correct modulation profile to support your average snr. also utilization as it increases it too will cause issues. I've seen 90% cause issues
Thanks for the help. But I would like to ask another question regarding dhcp-server and routing. In my dhcp-server I have a network 10.6.0.0/16 assign to the interface Bundle for CM get their config file, etc. and also I have 172.16.0.0/16 as secondary IP for Internet (in the border router I used NAT overload to the public IP). But now, we adquire a pool of /22 public IPv4. Is correct procedure to replace the secondary IP from the private network to the public network? And also do it in the dhcp-server? And in the border router use external BGP? Or Have I just need to change the NAT overload for 1 on 1 NAT?
Please let me know.
I have to ask why would you use a modem network and cpe network that supports two over 65K networks? the arp table must be crazy.
no need for something so large. I'm assuming you have less than 2k customers, no? asking because the ipv4 network is just a little over 1k.
simply add the new network as a secondary on your bundle, add route on your dhcp server. routing on on your border router depends what the next hop is looking for. what you currently use should be fine.
I agree with cmcaldas. Don't remove any ip's from your bundle yet. Simply add the new network as a secondary.
Here is an example from our lab cmts:
interface Bundle1
ip address 10.2.224.1 255.255.255.0 secondary
ip address 10.67.132.1 255.255.254.0 secondary
ip address 24.229.199.193 255.255.255.224 secondary
ip address 10.26.35.249 255.255.255.248 secondary
ip address 10.16.195.1 255.255.255.0 secondary
ip address 10.19.13.1 255.255.255.224
they are all small subnets, only because its for testing (and no, they aren't real, I just made the numbers up,) but the amounts come from a real cmts.
These scopes are for different things, voip, dsg, hosted-phone, cpe and modem.
Before you would remove any ip's from the cmts, you must disable the scope in your dhcp server first, and let all the ip's clear! If you don't, anything using an ip in the scope you are removing will simply stop working. Best to let dhcp migrate the devices to new scopes gracefully!
Thanks. I didn't knew I could add as much ip secondary address as I want. I will do it as you mention to avoid issues with the clients.
Thanks a lot for the advice, I was wondering how I should apply the new public IPv4 pool.
First, add the network into your bundle group. Make sure you add secondary at the end, or you will stop all traffic to/from the box!
Then, if need be, add the network to any acl's..
Make sure you can ping and traceroute to the ip address you've added to the cmts. If so, then it is safe to add it to your dhcp server, and it will start giving out ip's as needed.
ODG