0

I have a server running debian 11 server running Mailcow. It works fine, but every now and then one of my clients will report a connection error. This connection error only appears when they use their wifi connection - using mobile data temporarily solves the problem.

The exact wording of the connection error they get is dependent on the browser, but it's along the lines of

"this site can't be reached - xxxxxx.com took too long to respond. try checking the connection, checking the proxy and firewall. ERR_CONNECTION_TIMED_OUT"

OR

"This site can't be reached. xxxxx.com refused to connect. Try checking the connection. ERR_CONNECTION_REFUSED"

The wifi network is connected to the internet fine, the client can always access every other website on the net, just not my server.

There seems to be no pattern of ISP, or any other common factors I can identify. It just randomly happens to clients sometimes.

I tell them to reset their wifi router, and sometimes this works temporarily, and sometimes not at all.

Mailcow uses ipv6, but I recently disabled this and reverted to ipv4, thinking that might be the problem. Worked for a while, and then BAM, today I had yet another client complaining of the same issue.

Annoyingly, I can't reproduce this issue myself - my wifi, and every wifi network I've tried to connect form, works fine.

I've contacted my server hosting provider (binary lane), but they were useless, and gave me nothing at all.

Does anyone have any ideas of what might be causing this? It's very frustrating, and I can't find any ideas on the web of how to debug or fix this.

Improve this question
3
  • 1
    You can't diagnose your clients problems from your server side. Commented Oct 5, 2023 at 7:04
  • Ask your clients if your server is the only one they can't reach when the problem occurs. If other services are also affected it's not your problem. Commented Oct 5, 2023 at 7:05
  • Yes, it's definitely only my server they can't reach. They can reach all other websites, including their other email providers. Making my service look pretty dodgy. I know what you mean about diagnosing client problems on the server side, but it happens so often that it must be some common underlying cause. Commented Oct 5, 2023 at 9:50

2 Answers 2

Reset to default
1

Just to be clear, the error manifests on client devices which are connected to a remote wifi network you don't manage?

This connection error only appears when they use their wifi connection - using mobile data temporarily solves the problem

So you have a significant sample of data from users whom regularly use mobile data and those whom use a wired connections which proves that the issue only arises with Wifi? Or is it simply the case that most users are on Wifi most of the time? I suspect the latter. In which case it is very misleading to assume that the problem is wifi related.

The medium over which IP packets are carried beyond your router should have no impact (beyond the intrinsic congestion/bandwidth/packet loss) on the connections seen from inside your network. If the users cannot connect via their wifi connection but can using mobile data then there is a problem with their service provider, wifi network or device. The errors you've reported here suggest there is not a DNS issue.

Really, this is a problem of perception.

The best way I know of to address this perception is to provision a third party monitoring service (there are lots) to check on your service. There are lots to choose from, and many offer a free tier to get you started (I use uptimerobot). But for preference you want one where you can publish the availability.

From a technical point of view (and again based on the information supplied here) the clients seem to be having an issue with a specific network path - you can provision multiple network paths. The easiest way to do this is via a CDN. Again there's lots to choose from. Just make sure that you use one which can publish your endpoint on different subnets and provision these in your DNS.

Improve this answer
1
  • Perhaps I should have clarified - it was only a single user at a time reporting the connection issue, all my other clients could still connect fine. The client having the problem couldn’t connect using their static ip wifi network, but switching to mobile data they could connect fine. Turns out my server was blocking the client’s ip address Commented Oct 7, 2023 at 2:07
0

I figured it out - low IQ problems.

my mailcow server was blocking the client’s IP address with fail2ban.

The client’s computer was running outlook, and continuously trying to login with an incorrect password.

When he enabled a VPN, or used mobile data, the fresh IP address would connect for a while, then stop working.

Improve this answer

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.