No Internet Connection

[yeahman]

Hello,

I have 2 interconnected buildings with 35 APs total. About 300 wireless devices (mostly phones, some tablets) connected to the wifi system on a daily basis. My APs are all MR42 on the floors and MR52 in my boardrooms. I have 2 internet connections with different ISPs and the connections run active/active.

Over the past few months, and in only 1 of the buildings, my users have been reporting they can connect to the wifi, but then (in iOS) it will say "No Internet Connection" with that familiar orange exclamation mark. This happens across all SSIDs.

So, I don't know where to start looking here. Only 1 building affected. All devices run the same firmware and share the same internet. And the issue is intermittent and it *seems* like when it happens it hits a particular AP all at once..but this is only anecdotal i have no read evidence.

The fact it's only ever been reported from one building tells me it's something in this building...but I can't rule out perhaps a few wonky AP units that may be misrepresenting the issue here.

Any pointers on where to start looking?

Also, bonus question: Anybody know the exact mechanism iOS uses to determine internet connectivity? What specifically has to fail for iOS to report "No Internet Connection"? Is it a DNS request? an HTTP GET? Anybody know?

Thanks.

[Philip D'Ath]

Any chance the DHCP scope is running out of IP addresses?

[Philip D'Ath]

I think Apple devices try and connect to captive.apple.com to determine Internet access.

[Philip D'Ath]

Actually it looks like they try to connect to these three URLs:

captive.apple.com
www.airport.us
www.thinkdifferent.us

[yeahman]

Thanks Philip.

I'm not entirely sure. I have it set such that the AP handles the DHCP internally using "Meraki DHCP".

If their internal DHCP database was exhausted, I think that would show up in some logs perhaps? I haven't actually looked for that sort of thing. Sounds like i should do some reading on how Meraki DHCP works. Thanks for the lead!

[MRCUR]

Have you verified the Internet connection is up via a wired connection? Are the AP's checking in with Dashboard while this is happening?

[yeahman]

Thanks for your response. Yes, all 300 users use the same internet connection for various things. If the internet was down, I would definitely know about it.

[vassallon]

Are other devices able to connect to that AP when the device in question is having issues? It sounds more like a device specific issue. Is this iOS device up to date for iOS?

Have you tried resetting network settings on the iOS device when it is having issues to see if that fixes the problem?

[korebell]

This sounds like the perfect use case for Wireless Health. This should help you quickly identify the root cause Association, Authentication, DHCP, DNS, etc. It will also help you identify if certain APs are experiencing more issues than others as well as different device manufacturers. This feature should have been enabled on your dashboard within the last week or so.

Here is the document: https://documentation.meraki.com/MR/Wireless_Health

[yeahman]

Thanks every for their replies so far.

I have some more info, as I have been able to work on this a little more today.

I ran into this problem...and identified the AP. I was getting "No Internet Connection" and noticed I was associating with the AP but the DHCP address was not being assigned (stuck on the 169.x auto-config address).

I rebooted the AP remotely. As soon as the AP came back up, I connected right away and 3 other iOS devices in the vicinity also connected right away. Previous to this, only a single android device was connected for 2 hours.

So...i'm using "Meraki DHCP" on these...which I assume is a dhcp service on the AP unit itself. Is there any way I can dig into the logs of this dhcp service or the AP?

[MRCUR]

There are no visible logs for the DHCP service when running in NAT mode. If you contact Support, they may be able to see a bit more info (or at least check AP load CPU/RAM) around the times you're having the problem.

[cta102]

One issue a customer suffered for was exhausting the DHCP pool because the (default) 1 day lease period was left in place.

Reducing that to one hour took care of the issue, but they did have a very large numbers of users passing through.

Do you think there would be enough devices passing through that area to create a large number of DHCP entries?

It's only a suggestion as I haven't used the AP based DHCP so don't have a clue what it's default lease time is (but it's easy enough to check the least details from a connected device)

[yeahman]

I'll check on that. Thanks for the suggestion!

I don't imagine there are that many devices passing through, but definitely worth checking out.

This is the point where I say the inevitable regarding Meraki products...I sure wish there was a way for me to get down into the weeds of the system when I need to. Would save alot of time going back and forth with support.

[MRCUR]

---

@cta102 wrote:

It's only a suggestion as I haven't used the AP based DHCP so don't have a clue what it's default lease time is (but it's easy enough to check the least details from a connected device)

---

You cannot configure the DHCP options for NAT based SSID's. The AP's use a /8 mask for the NAT subnet so it's very unlikely the scope is running out of addresses (although it's possible there's a hardcoded limit but I still don't believe that's the issue).

[cta102]

Without knowing what size the table to hold the clients is permitted to grow to it's purely speculation.

However with short(sih) lease times it's incredibly unlikely but with 24 hour leases it's a possibility (was seeing sites with 2 APs exhausting /21 DHCP pools within about 5 to 6 hours.)

Although I would consider it unlikely it's just about valid enough to feel irritated that I wouldn't be able to view the lease table and discount the possibility.

We once had an issue with a particular Aruba firmware release where the AP got upset at 96 clients (it was such a nice number I started laughing about a mistyped a mask value) then the firmware was pulled, happily that only caused issues on a couple of sites.

Edit: The Wireless Health tab looks like it would have been pretty handy for this one though.