Help needed with intermittent internet

You start by saying that there are intermittent outages, but you only specifically mention it being down first thing in the morning. Are there on-going problems at random times during the day, or is this only happening when the connection has been idle for some time?

If the problem appears often after the use of the Internet connection has been unused for a period of time, then I would check to see if the server is shutting down the NIC because of some power management setting.

Also, if it only happens overnight, set it up to run ping continuously overnight, and see if the problem goes away.

Something else you could try is instead of having the Novell server connected to the Internet, connect a different system. Of course this wouldn't be practical during the work day, so you'd probably have to do this over the weekend. Make sure there are no power management settings to shut parts of that system down when idle, and test to see if the Internet connection goes down.

BTW... You don't mention how you get the connection back up when it goes down. Are you rebooting the server? Power-cycling the modem? Calling Comcast, and having them reset the modem remotely? What does, and what doesn't work to bring the connection back up?

Thank you very much for your excellent reply. Here are some answers (I'll get other answers this evening).

It happens throughout the day, mostly on the server's static address (for terminology clarification, I'll use "gateway" address and "server" address from now on). This fact led Comcast to conclude pretty early on that the problem was on our side, since when we called during the day they could ping the gateway address. That was reasonable, enough, until I started finding the gateway address down at certain times, too, and often it's down in the morning. However, over the weekend, I had failed ping tests to the gateway at various times, day and night. Another fact perhaps worth mentioning is that every workstation in the business is turned off at night and weekends. So whatever is happening along the lines of Comcast's theory is limited solely to server activity. In other words, if some activity on the lan side indeed is "locking up" the Netgear, it's emanating from the server, not a workstation, i.e., a malware-infected computer flooding the network. (and I'm being open-minded, but I have my doubts in any event that something lan-side could indeed lock up a cable modem). In general, there seems to be no relationship to idle time.

This does not appear to be the case.

As mentioned, it happens day and night.

That is my plan for tonight. I'm going down there in a little while to connect a newly-formatted PC to the connection. That should be very informative.

That is another strange aspect to the story. For 3 straight days, the modem began working again during my call to Comcast. I would call in, and after identifying the customer the Comcast rep would attempt to log in to the Netgear to "take a look." The router would start working immediately. In fact, during that 3-4 day stretch, one of the reps and I used the "sleep mode" analogy, such as you mentioned with the power management suggestions. However, many Comcast folks have assured me the Netgear has no sleep mode. This morning when I called in, the Comcast rep offered to reset the modem, which brought the internet back up. When I said that indicated a problem on their end (or with their equipment), she reiterated Comcast's position that since they've swapped the unit out twice the problem MUST be on the client side. They think something is locking the Netgear up, requiring a reset sometimes, while other times not.

jm

I'd start with a separate router and modem setup if possible. Should make it easier to isolate your problem and it might just go away.

Funny you should suggest that, as I just returned from driving halfway to the customer site to install a Linksys. Instead, I turned around, deciding to have them turn off the server tonight to see if the gateway still "locks up" or otherwise stops responding anytime between now and 7:30 a.m tomorrow morning. I have a ping test monitoring the gateway right now. If the gateway still has problems with the server and all workstations turned off, I think Comcast's theory is pretty much debunked.

Agree?

jm

After further consideration, I don't see much value in this bit of troubleshooting. Unless the unit is bad in a manufacturer defect kind of way (which I seriously doubt, since we're on our third one), there is no way the unit will fail with no traffic going through it.

As $Bill suggested, perhaps it would have been better for me to go ahead and install the Linksys and remove the server-as-internet-gateway from the equation.

jm

It would go a long way towards it if indeed all the LAN equipment is off, but that really doesn't solve anything.

Next try sticking that Linksys in and reconfiguring the Netgear as a plain router (I assume that can be done) or replace it with a another router.

The Netgear is provided by Comcast. It's a combo router/cable modem, and it's the only device they offer for business use with static IP.

jm

Borrow one or buy your own for $40 and sell it later if need be. At least you'll have a chance of finding the problem for $40 (or less if you pass it on when you're done) and I assume you already own the Linksys.

If it works with your equipment, they won't be able to pass the buck.

I asked about this a few days ago, and they told me they do not allow customer-owned cable modems. I pressed the issue with another csr, who insisted they cannot do this, since there is no way for them to provision the service on equipment other than theirs. She asked me, "How in the world would we configure it?"

jm

It just dawned on me that you didn't say wether your Linksys was a modem or router - I had assumed modem.

That's got to be both the funniest and stupidest thing I've heard lately. Who trains these people ?

You should have told her the same way every other ISP does it. :)

I don't think she meant that it wasn't possible in the global sense, but rather it isn't something that they're allowed to do, and there aren't any work-arounds, either.

I don't know why Comcast requires that those business-class customers use only the modem that they provide, or why they chose that particular modem. I suspect it has to do with an SLA, and their need to minimize the variables out of their control.

But it's not a training issue. The company could provision a customer-owned modem, but they've chosen not to allow the agents that ability. So they, the agents, don't have any way "in the world" to provision it, but that's probably not the best way she could have gotten that point across.

BTW, this brings up another point. Is there an SLA in the contract? The cost of business-class service has gone down considerably in most markets, and that may be because they only include priority support, and not an SLA these days But I suspect that if there is an SLA, it would have something to do with the modem being up, but not necessarily anything beyond. If there's an SLA, they would probably keep records to protect their liability, and those records might also be available to the agent. I'd ask what they show. And if there is no SLA, and they don't keep those kinds of records, there ought to still be a way to escalate the issue to the NOC, and have them do such monitoring.

It would be hard for a customer to do the monitoring themselves. A DOCSIS cable modem has an IP address on it's WAN side. Normally it's a class A private range IP address, so monitoring the modem would require being on their network. And it would also require knowing what the modem's IP address is, and that's not something that there is any way for the customer to discover. Perhaps the CRS is able to see what that IP address is, but it's also possible the tools they have hide that address as well.

I'm also curious about what the modem's indicator lights are indicating during the problems. And I also wonder what the modem logs show. Confirming that it's a TCP/IP problem, and not an RF problem (or vice versa) is an important step that should have already been taken. (And probably was, but, hey, if we're brainstorming...)

The activities of the front panel lights don't tell me anything, although I'm not sure I would know what I'm looking at anyway. The power light is on, of course, as the "network" light (as the csr called it), indicating sync with the ISP. Then there two opposing "lightening bolt" lights (upstream/downstream) that flicker contstantly. They seem to flicker 3 times in unison, and then several times alternating. Then there are numbered lights for the lan ports connected.

As for network monitoring, I asked about that, and, you're correct, it's not something they can do from their local help desk. They simply don't have the tools. I had another support group on the phone on Saturday, and she actually referred to the "real" IP address, not our static (I wrote it down if that would do any good). They could not monitor the network traffic, either. Interestingly, she said her group could not even log in to the Netgears. Comcast has recently purchased Time Warner - our old ISP - and Adelphia. There seems to still be a lot of fragmentation.

The question remains: Why is the Netgear locking up? Some Comcast reps still deny their equipment is ever going off line. But while I was at the customer site today I proved it again, at least to my satisfaction. The internet went down, I reset the Netgear, the internet came up.

While Comcast's basic stance is that the problem is on our side (after all, they've tried 3 Netgears), one event seems to stop them all in their tracks when I bring it up: Last Thursday during an outage, I called Comcast - as has been my habit every day for 3 weeks - and the rep was in the process of logging into the Netgear when the Netgear did a full power reset. For some reason, this didn't flip my lid. I simply asked the Comcast rep, "What did you just do?" He answered, "What do you mean?" When I told him the Netgear powered off and back on, he didn't hesitate: He said, "That's never supposed to happen. I'll have a tech out shortly." Thus, our 3rd Netgear.

With that in mind, what in the world are my options?

thanks to EVERYONE who has given of their experience and time in helping brainstorm this issue.

jm

When they replace the modem, are they just replacing the modem, or are they also replacing the power cord and power supply brick? If they just swapped the modem itself, the problem piece may be still sitting there.

And what is the power supply plugged into? If it's plugged into a power strip, surge protector, or UPS, try taking that out of the mix. Plug the power supply for the modem directly into a normal wall outlet. Use a heavy-duty extension cord if there isn't one close enough.

In fact, if you can reach an outlet on a different circuit, try that -- especially if there's a laser printer (or any other appliance that periodically needs to heat-up) plugged into the circuit it's on now.

Your modem is receiving radio waves. Just like any radio, signal integrity is determined by signal strength, or more important, Signal to Noise ratio.

Many cable modems for some reason cannot bother to provide that critical parameter. Without it, then one can only speculate whether the cable connection is good or bad. As obvious from responses - modem 'swappers' are only speculating. Wildly replacing modems without first learning what is wrong. Without that 'signal to noise' ratio number, then no one can even know which side of the modem is failing.

That number is displayed by a modem on a status page. But only if the cable provider had an engineer (not a bean counter) selecting modems.

Meanwhile, eliminate other reasons for intermittent operation. Start with something so often forgotten by untrained cable installers and that can permit household or neighborhood appliances to create intermittent failures. That cable must drop down to be connected 'less than 10 feet' to the same earthing electrode that is also 'less than 10 feet' from the breaker box. Make that earthing connection before cable enters the building. Connected to an earth ground that every other utility also connects to; using a ground block (as even sold in Lowes for $2) and 'less than 10 feet' of 12 AWG wire.

Next, where does that cable wire route. What else connects to it? An alternative test is to route a wire direct to where cable enters the building with nothing else connected - and change nothing else. Does system work more reliable? You would know that immediately if modem provided S/N ratio numbers. Of course this test performed only because useful information to identify a fault is (apparently) not available AND your cable 'tech support' has no idea.

Meanwhile, a computer can connect directly to modem's status page - constantly. That being one way to monitor customer side of the modem. When outside connection is lost, is that modem status page still accessible?

Of course, anything you do to make the problem worse should help find the failure. Find a failure before trying to fix or replace anything. Since they did not do that and then replaced modems, what do you know? Nothing. You don't even know which side of the modem is problematic because they did not identify a problem before fixing it. Instead they shotgunned. Therefore nothing useful was learned. Therefore zero progress has been made. You don't even know what is good. Everything is still unknown.

If available, get that S/N ratio number that every miminally acceptable modem (cable, DSL, etc) must report.

Thank you for your reply.

I cannot access the modem's config, but Comcast (both locally and the level

2 guys in their Denver NOC) insist that their levels (S/N, upstream, downstream) are all "within spec." And, regarding their support, I've now spoken to more knowledgeable, more sincere Comcast people, and I'm convinced for the most part that they are simply as stumped as I am about this whole deal.

How strange is this: Every morning the internet is down. I can neither access the lan nor ping the modem's gateway address remotely during this time. However, as soon as I call Comcast and the support person pulls up the portal to "take a look," the internet starts working again.

It's the most bizarre thing. It's as if the modem is going to sleep at night. However, in Comcast's software the modem shows to be online the entire time.

jm

Comcast on their end only sees their connection to line amplifiers. That number only says whether everyone is getting connected. It says zero about a connection only on your end.

Again, your problem is classic of intermittents which is why a provider should provide S/N ratios on your modem that you can monitor. Did the Comcast tech look at signal strength from modem while standing in the room - or instead only with his measuring equipment? Modem can report valid and useful number. Their equipment can only report what modem 'might' see - and not from their equipment on other side of amplifiers.

Remember a trend among techs. It was never a problem elsewhere; therefore it is not your problem.

Listed were numerous things to perform. Did they confirm above listed connections? Your post did not reply to what can create a significant problem and can only be identified by visual inspection.. And yes, many techs do not appreciate the engineering even in simple earthing. Often because it did not cause a problem elsewhere; therefore is not your problem. Did they perform the 'reroute' test? If not, then why not? Is existing cable good only because they know it must be good?

Meanwhile, what did you do to verify a problem does not exist on subscriber side of modem? For example, as your computer pings the modem constantly, what happens to ping (the numbers) both when connections are working and not?.

Again, their levels can always look good. But only valid number is S/N number read directly from your modem. Any only valid number is both when connection is working and has failed. That number must be measured on your end of wire and is only useful if read directly by modem. That number read anywhere else says nothing about your unique connection.

Your post implies a "we have d> I cannot access the modem's config, but Comcast (both locally and the level

Not true. At any given moment, they can tell how many modems are up (and down) on each node, and which ones they are. This is at the IP level.

Again, not so. Their monitoring equipment can tell what each modem is reporting.

What the agents on the phone see is what the modem is reporting. It is not a measurement from their end. It is the numbers that the modem itself reports back.

This whole idea that it has to be a grounding problem is quite a long-shot, and really doesn't fit with the symptoms reported to this point. However since we're all starting to run out of ideas, and practical isolation troubleshooting hasn't been very fruitful, he might as well start checking random long-shots.

I also don't detect any bad attitude on his part. What I see is that he tried everything he could think of (which is different from everything), and still has not found the problem, and is therefore trying to backtrack, and look for things that may have missed. Hardly the "bad attitude" you're attributing to him. If anything, his attitude is better than most of the people who post problems here.

I'm no expert on internet/WAN, by any means, but I don't believe you are correct on this point. Even the local first-tier support people can log directly into the modem using what I believe they call PSV, peruse the settings, readings, config, etc. Granted, the information they glean appears to be rather limited - compared to the NOCs of other providers with whom I've worked - but I think you are describing Comcast's support as much more "in the dark" than they actually are.

I'm not following the reasoning.

I respectfully disagree with this generalization. I have been a technician for seven years, and I used to operate a team of 13 telecom technicians. In general, technicians try to solve problems to the best of their ability. Your characterization is the exception, in my experience.

Several issues here. First of all, I have not had time to distill your suggestions and put them into context. Secondly, Comcast is not simpy going to execute every list of suggestions that might come their way. Third, while I sincerely appreciate your willingness to help, and I respect your expertise, I do not believe your earthing theory is valid. The reason is that cable internet has been functioning perfectly in this location for years. Only within the past 3 weeks have the problems arisen. While I'm no electrician, I do understand the basics of electricity and grounding, and I do not see how this factor bears on an internet connection's reliability after working so well in the past. However, I may be missing something, and I welcome your arguments to the contrary.

This is a valid question, and there is much to say here when I have a little more time, but for now I want to ask: What are some examples of problems on the subscriber side might cause intermittent internet?

To some degree. However, more accurately, the attitude is: "Yes, something is wrong. We have caught our modem offline a time or two. However, for the most part our equipment shows your modem to be online. Whatever is causing the Netgear to stop responding is a result of a problem on your LAN side."

They do not disagree that something is affecting the performance of their Netgear modem. They just don't think the problem is "their fault."

Without re-reading all my posts on this issue, I'm not sure what details of included and left out. However, I can say with certainty that everyone involved is "looking for a problem." They/we may not be looking in the way you would look, but that doesn't negate the effort.

And please clarify your suggestion regarding recording ping results from subscriber side when connections are good and bad. Ping what? When the internet is down, the Netgear [usually but not always] will not respond to a ping, either from within or from the outside. Perhaps I've missed your point.

Thank you for your determination to help.

jm