Internet Drops Randomly When Using Procurve 2910

T

timberdoodle

Gawd
Joined
Sep 22, 2008
Messages
878
So we recently moved our Comcast demarc to our new server room which resulted in moving the SMC modem from our old server room where it directly connects to the firewall - to a new room where it connects to a 2910 switch, LACP to our L3 switch, LACP to another 2910, then to Sonicwall.

Since the move our internet has been dropping off randomly for about 2-5 minutes then it will come back online. This happens anywhere from 10 to 50 times a day. Sometimes it happens a half hour apart, sometimes it is 15 minutes apart, sometimes it will go 8 hours without issue.

The modem has been replaced, the Comcast line has been "tested."

The modem remains online when we lose internet connectivity - I can connect to it using a laptop on it's lan port and ping out to google.com while an computer connected beyond the switch and sonicwall cannot connect.

I have run a constant ping from a home comptuer to the modem as well as to our sonicwall (both static addresses) and they both seem to go offline when we lose internet connectivity.

Since the issue started, I moved the Sonicwall so that it connects directly to the 2910 that the modem connects to, put it in a vlan with the modem so they only talk to each other and it still results in the same intermittent disconnects. I checked all the port statistics on both ports as well as the event logs, and trunk links and find nothing abnormal.

I take the switch out entirely and plug the modem directly into the firewall like it was prior to the whole move. Works fine.

Traffic is not an issue and the switch in question also connects all my production servers and iSCSI SAN together without issue.

Thoughts?
 
Is that SMC's WAN port 100Mbit or Gig? Configure the port on the 2910 that the modem interfaces with to 100FDx or auto-1000, which ever is applicable. Anything but auto/auto. Same goes for the port your Sonicwall now interfaces with.

From inside your network ping the modem management interface, 192.168.100.1 (usually modems use this gw, unless this network exists elsewhere on your network)

Second try pinging the gateway of your public ip range.


Edit: Don't forget to try swapping the ethernet jumper between the modem and switch, sonicwall and switch just in case.
 
Last edited:
timberdoodle said:
Traffic is not an issue and the switch in question also connects all my production servers and iSCSI SAN together without issue.

Thoughts?
Click to expand...

Are you having spikes in iSCSI traffic during the internet outages? I would check the CPU/MEM on the switches during the outages and make sure they are not spiking.
 
Thanks for the reply guys. The SMC LAN ports are all 1000FD, the switch was set to auto at the time. The Sonicwall is also 1000FD with auto config on the procurve side. The WAN port is a coax cable so I can't exactly do anything with it.

There is a possibility that my veeam backups overlap the outages in some cases but not all. But even if they do overlap - the server I backup to only has 1 nic port presently.

Edit: When I moved the firewall closer physically I also put in new cables. Same issue.
 
Is it on it's own VLAN on the switches? So you basically have a VLAN between just the modem and sonicwall, correct?
 
/usr/home said:
Is it on it's own VLAN on the switches? So you basically have a VLAN between just the modem and sonicwall, correct?
Click to expand...

Yes that is correct. The LAN traffic gets routed to the 2910 from our 5400 then out to the sonicwall X0 port which is all on one vlan. Then another vlan just on the 2910 takes the interface from the modem and goes right to the X3 interface on our sonicwall.
 
timberdoodle said:
Thanks for the reply guys. The SMC LAN ports are all 1000FD, the switch was set to auto at the time. The Sonicwall is also 1000FD with auto config on the procurve side. The WAN port is a coax cable so I can't exactly do anything with it.

There is a possibility that my veeam backups overlap the outages in some cases but not all. But even if they do overlap - the server I backup to only has 1 nic port presently.

Edit: When I moved the firewall closer physically I also put in new cables. Same issue.
Click to expand...

What Im saying is lock down the port speed between devices where you can. I recently had a customer have weird intermittent connectivity issues between his sb6121 modem and a Watchgaurd firewall. Had him set the WAN interface on the Watchgaurd from auto to 1000 full ans the problem went away.

The ethernet port on the modem is of course unmanaged so you cant do anything there. The coax on the modem is simply a RF bridge back to the the CMTS, so theres no true "port speed" really happening there (for the sake of this conversation)
 
Last edited:
I can try that but duplex and/or portspeed mismatches should be logged in the switch events. There were none for those ports.
 
Even if they aren't logged you could still have some issues. I hardcode all trunks, internet connections, servers, etc just in case.
 
The afformentioned customer was logging all port stats and never noticed anything in the logs.
 
Related question on the port config.....

MDI mode, do you set that and if so what should it be. I see it is all set to auto but on this particular switch with the 4 trunk ports, 3 set to mdix, 1 set to mdi. We use straight through cables for everything.
 
Ok I went in this morning to configure the port speeds. When I set the ports to 1000-full on any port on the procurve it gives me an error message that the value is not available on the selected port. I was only able to set auto-1000 so the speed is set but the duplex is negotiated. Another strange thing is when I set the speed on the SMC modem to 1000 full it would go offline for about a minute, then come back up with 100 full set. I'm not sure if it can't use 1000 or what but I figured our internet service is slower than that anyways so I set it and the switch to 100-full which took in both places.

I set the firewall to 1000 full and set that port to auto-1000 for both the wan and the lan ports. We'll see what happens!
 
Gigabit can only be full duplex. Some switches will let you hardcode gigabit, some won't.
 
Any reason why you are not wired direct to sonic wall from cable modem? I understand they are in different server rooms now, but do you not have a way to patch a cable directly over to it? Seems odd to have 3 switches in between your ISP and WAN edge device. I am sure your bottleneck is in between the two. Can you not take down the sonic wall during a maintenance windo and move it into the same room as the modem?
 
Also, try to isolated your issue a bit further by connecting a machine to the first switch and do an extended ping to the WAN GW...or DNS server etc....and monitor for drops. If you find no drops off the first switch, then move upstream from there to the layer 3 switch and do the same. You may find the drops are occuring deeper into your network as oppose to the edge. Food for thought
 
No it is in the same room now, one hop from the modem. As I said I had moved it since the problems started. I plug it into a switch for flexibility such as span port in the future.

Edit: I noticed there were other VLANS tagged on the ports in question even though these ports are untagged for the wan vlan. Not sure if that has any impact but I removed any other tagging and no drops since...
 
So if you hook a computer up straight to the modem and do a constant ping, will it ever drop?
 
I can't say for certain. Sometimes the ping would drop but I wasn't sure if that was just Comcast doing tests or not. Cause other times we would lose Internet but I could still log into the smc and ping google while my laptop behind the firewall could not
 
I work for Comcast. You wouldnt drop because of us doing any tests unless you called customer support and had them "hit" your modem. If you can lock up on the fwd/rtn, got link back to the CMTS, and you can ping the WAN GW...you are good to go.

If you are connected direct off the SMC and attempt to ping out and drop packets....then there could be an issue with the circuit itself. You stated this is not the case.

Sorry but I am still a bit confused on your setup....correct me if I am wrong...

Comcast SMC---->Sonicwall WAN port---->Sonicwall LAN port---->production LAN??

Maybe check the sonicwall logs to see if there is anything funky going on and also double check your WAN settings within...make sure it all matches up according to design. If you still feel it is Comcast causing the problems, send me a PM. I can give you a customer support number that is "VP" if you will. Sends you right to the top with priority.
 
Well let me just elaborate and explain that we had the Comcast tier 2 out a few times as they switched our line from the street overto a new line due to construction. They tested the line at that time without issues. I have been in touch with tier 1 and tier 2 since monday regarding the disconnects. I havent had issues since earlier today but if I still have trouble tomorrow I will take you up on that offer!
 
We have not had any drops since I removed that second vlan tagging on these ports. I thought once a port was untagged in a vlan no other tagging took place. Sloppy work on my part I guess. Thanks for the help guys.
 
The wan vlan including the Sonicwall and modem ports was untagged or access ports. There was also a separate vlan that was tagged on those two ports.
 
You must log in or register to reply here.
Back
Top