 I think there is a question in here in the chat for you, Karan. It says ICMP based monitoring for website does not give accurate results. So yeah, you have to take on that. Yeah, that's very true. And like I mentioned in my blog, some of the upstreams even block ICMP like even github.com, Netflix.com if you try to monitor them, they'll just outright block your ICMP. And also if your website is behind a CDN, you're actually just monitoring the CDN end point rather than your own end point because a lot of these CDNs, they won't forward your ICMP packets to the upstream. What mattered for me wasn't monitoring the end point itself. So for that, I'll just do a layer seven, HTTP based monitoring. For me, what mattered was just to get a general overall picture about the network health and the sites that I basically frequently visit and I know they are spread across regions all over the world. So in that way, I just get a overall picture of how my ICMP is performing. It's not really a particular end point of monitoring. For that, you should use HTTP or other metrics if you can. And this just tells me, okay, some problem is there. Then if I need to dig in, I can basically do a trace route and or basically do other kind of metric monitoring because like has it had a webinar on write probes as well, atlas probes as well. So even that gives a lot of useful information. So that helps to give us those problems. Cool. So there is this another question from Brava. So it says, so I may have heard it wrong, but did you mention using Prometheus and influx both part of the stack? If so, was influx being used only as a data store for Prometheus? I can give my date. So I'm not using Prometheus at all for my use case. I'm using influx TV, like only influx TV. There's no Prometheus in my entire setup. What about you Karan? Yeah, so I don't have influx TV at all in my stack. What I have is just probes, Grafana and the exporters which are running just as a stateless process in my Raspberry Pi. So like I mentioned, telegraph seemed very appealing to me because it could just outright export the matrix in Prometheus format. So I just use Prometheus as a timescale DB. So I really don't need to use influx DB in my current stack. And how it's architected is the Raspberry Pi runs the exporter which is connected to my home internet using a LAN cable. And I use something called a stale scale which is a mesh network based on wire guard VPN. And I have a digital ocean of like small $5 instance running in Bangalore. So the Prometheus is connected. Prometheus is basically listening on tail skin interface and my telegraph exporter is also listening on tail scale interface. So that's how they just talk to each other. So yeah, I mean I don't really have influx DB in my current stack. Yeah, I guess it speaks a little about telegraph, right? It's kind of flexible. You can use it with either influx DB or with Prometheus. The reason I went with influx DB was because influx DB integrates really well with a couple of other monitoring I'm doing. I'm also monitoring using SNMP. So I'm using the SNMP plugin in telegraph. And the entire stack for doing SNMP monitoring was kind of very easy. So I found that using telegraph and influx DB you could do both external as well as your internal monitoring. So like I was mentioning about the Grafana dashboard, you can see the DNS. This is also a telegraph input plugin. And it's just a metric which tells you the response time for your DNS queries. There's a DNS error also which will tell you if any UDP errors are coming in your DNS server. Other than that, there's a very simple uptime metric for the endpoint as well, which just subtracts the packet loss from 100, which will just give you a general idea that how much packet loss percentage your endpoint is having. And this doesn't have to be, you just have to be careful that you just use it for a general curiosity or like a monitoring purpose, not for the actual endpoint monitoring because like someone mentioned in the chat also, a ping is not a reliable way to monitor this. Yeah, all the other metrics, they are pretty basic. This is my main Bangalore VPS. And all the ping response time metrics are just plotted over here. Cool. I kind of have a question. So are you also planning on monitoring your internal network or just like the external stuff? Internal network, like using like you mean all the services that I'm hosting, are they up or not like that? No, asking your own network. Okay, so you mean the TealScale machine network or like Raspberry Pi and okay, okay. So I haven't really digged into that right now, but I mean, yeah, that sounds interesting because what happens for me is TealScale just takes care of all the part where you just connect your devices and it's pretty cool in a way that I can just visit a container which is running in my laptop. I can just listen on TealScale interface and just put the address in in my mobile and since mobile is also connected to TealScale, they both can talk to each other, but I haven't really looked into how I can sort of put monitoring into it because I think I'll need some kind of deeper like layer 4 kind of metrics running through the network over there. But I do have a basic like node exporter running in both Raspberry Pi and digital solution. Yeah, so I think since we are talking about home networks, right, like I can probably share how I'm monitoring my home network. So I'm basically doing the same thing as you are, running telegraph on a Raspberry Pi instance, but I've also enabled ICMP metrics for my router. So that displays all like a lot of metrics which you see inside your router as well and you can use that for like you can monitor your home network using whatever your router sees right. So that was like a previous version where I was trying to figure out how to do this and the initial router firmware that came with my router did not support ICMP. So I just flashed like one of these open source router formers and that basically helped me export all of that information. And after that, I can invest a little bit more money and switch to Ubiquiti, like using Ubiquiti Gateway and Ubiquiti AP. The interface for Ubiquiti is really nice. Like by default, it gives you a lot of information in there by default. But if you want to integrate that with your entire Grafana dashboard, you can do that also like they're already ready made, unified dashboard presence. So you just have to go to the Grafana website and just download one of those dashboards. Maybe I can quickly like share my dashboard over here for my internal network as well. Share my screen, application window, Chrome, share. Can you people see my screen? Let me know if you can see. Yeah, yeah, we can. Yeah, so I mean none of this is created by me. I'm just leveraging the stuff which has been done already by other people. But like investing in kind of pro-sumer hardware, it gives you kind of feature rich feature rich stuff which you can already leverage from other people's work, right? Like all I had to do was set up Influx TV on my Raspberry Pi instance. And yeah, and this dashboard is not created by me. So I can already monitor a lot of stuff on my internal network itself. So there's a lot of information which has been shared by default without having to do anything. So this is basically just a plug and play operation now because I didn't have to like do any of the stuff. So yeah. That sounds cool. And even Ubiquiti, they have a new retailer in India as well, right? Yeah, that's true. So Ubiquiti had a new retailer in India and they started selling all of their other locally also. It's a little bit expensive, but the convenience factor and the nice new one you get in return, I think for my use case, it was worth it. So I kind of switched over to Ubiquiti stuff. And yeah, let's say you want to, you're running probes or you are probably interested in leveraging the Atlas probes and plugging that data into your, let's say dashboards, right? So there is a Prometheus exporter called as Atlas exporter, which you can probably explore. And so it's a cool thing because it gives you the power of running a streaming or a live measurement into your dashboard. And the same guy who has written this in Go also has a ping exporter. So I highly recommend that you check that out as well. So this gives a, this kind of opens up a lot of possibilities. So in my case, I am also using Prometheus and using influx DB as well as using the Atlas exporter with a streaming measurement, which tells me latency from my network towards or latency from specific networks in India towards, let's say, the root servers, right? And that's pretty cool because that tells you how the root servers are behaving at any point of time. And it also tells you how much of traffic is staying within India and how much of traffic is kind of transiting outside of India. So for people who may not know about DNS root servers, so we have a lot of instances of root servers within India, IPv4 as well as IPv6, especially the F root servers say, for example, but, and even the K root server with the IPNCC. But number of times I have been monitoring the fact that traffic will not hit the local instance, the traffic will go outside India. So this becomes an excellent way of having a historical understanding of what is happening there. So that could be something that people can explore. Yeah, that's very interesting and nice input. I hadn't thought about that, like integrating that with Prometheus. But yeah, I think Hasgi calls have done a previous session on Atlas and ONI probes. So yeah, if people are interested, they can go check out that stuff. The other thing probably what I'll add to the mix is I'm just assuming that you have multiple internet connections like I do is monitoring on that front as well. So if you have, let's say, two internet connections, how about you folks monitoring both the internet connections? What kind of hardware are you using, if any, and what tweaks to the setup to do that? So the Unify Gateway, which I have, it has two LAN inputs. So theoretically, I could connect another one, but I've just decided to YOLO and I'm on like only a primary single ISP. The other one is just my 4G network, which I'm using. But yeah, this is something which even I'm interested in. So if there are any thoughts, please let me know also because that will help me figure out how to integrate both two different ISPs into a single network. So I have been thinking about getting a secondary backup connection, especially after yesterday since my internet has been down for like 30 hours. So what I was thinking was self posting just a DHCP server and integrating alert manager alerts where I monitor either of the networks and kind of switch the endpoints inside my Raspberry Pi. So I want to use my Raspberry Pi as a router and connect to both the internet routers using that and just depend on Raspberry Pi for my routing use case. So yeah, so in my case, I have multiple connections and what I'm using is basically a dual LAN setup. It's something that I picked up sometime back and I use policy routing primarily if you haven't explored this. So policy routing allows you to have policies where you could save. So I have multiple probes. So I have two internet connections and I have two different probes, software probes behind those two connections. Now, what I do not want is, say for example, if primary internet connection goes down, I don't want all traffic to move to the secondary because I want this first probe to go offline because that's what it is configured with. So using something like, I mean, there are many ways to do it, but one simple way of doing it or effectively, which I could think of is using policy routing and that kind of solves that problem. So while the first connection primary goes down, the first probe will also go down, but at the same time, the rest of my devices will switch to the secondary and vice versa. So that is something that you could explore. So the ubiquity devices, the USG or the edge router, you could do that there. You could use PF sense, whatever it is, you could take a mix of the six box, put an additional nick in it and configure that as a router. You could use open BSD or even free BSD and do it. I choose whatever works for you the best. Regarding this split, splitting traffic. So even I tend to do that, but using DNS. So I run a local code DNS server and I basically whatever the DNS zones that I want to set off particular DNS zones that I want to white list, I just split my traffic based on that. So maybe you can also check that code DNS as well. So I had a question for you. You were talking about using BSD and BSM and stuff like that, right? Is there any like ready-made easy to roll out solution for using two different ISPs? Not in a like active passive way, but in like kind of a load balance tree or something because the problem which I think I would run into is like having those sticky cookies. If some of my packets are going to one ISP and some of my other packets are going to the other ISP in a load balanced way, then it will kind of break my sessions, right? Right. So one BSD option which is probably has a UI is PF Sense. So PF Sense is something that people can explore if they just want instead of using policy routing, if they just want to switch between multiple connections, that could be one option. There are many others. I mean, if you just search for alternatives, you'll find many. On your specific question, the last part of the question where some part of the traffic is going this way, I mean one through the first connection and some part is going, I mean depends on what you're trying to do. So in my case, using policy routing, I have a lot of things in there where I have certain, so I have a lot of devices. In fact, it kind of reminds me that I need to get a rack, a full-fledged, full-blown rack because my home network is an enterprise network right now, where I'm kind of looking at p-caps, taking all the traffic using a span port and doing a lot of things there. But as an example, I have certain devices which I always want that traffic for those devices should go through, let's say, the primary internet connection. And there is the other part of the rest of the devices, like say in my home network, my family, whatever devices they are using, although I'm at home, I don't want them to pester me with respect to that the internet is down all the time. The network NOC is right besides in the next room. So I primarily don't want to fiddle with any sort of meddling with what happens to those devices. So there is a free flow with respect to those specific devices where the primary fails, everything switches to the secondary and if secondary fails or whatever vice versa comes back to the back. That's what I have in place. I hope that answers your question. What if I do not want to roll out my own hardware? Are there like any ready made products which would integrate two different ISPs into a single home network? I guess it would be interesting for other people out here. I don't think that would be seamless and possible without having a hardware in place because the logic is being done in the hardware device. It could be an x86 box, or even there are other things which I forgot to mention, which one people can explore. There are a lot of these routers, let's say TP-Link as a very common popular brand in India. So there are a lot of these TP-Link routers which you can put open WRT or open source formverse like what you were talking about earlier. You could configure the multi-wan setup using the most economical TP-Link router. But the problem with that is having a compatible router which works with the open WRT or any other open source formverse is always a question mark. In my case, I have also explored that. So if you go to the open WRT or the Wiki page, I'll put the link in the chat as well. So they have a page where they list the hardware compatibility. So if you search for TP-Link, it tells you which one is compatible, which model and so on. So you need to be very picky about it. So in my case, it was me calling up the distributor and saying, okay, I need this specific version and I'll pay you the money, but before you ship it, please send me a picture of it so that I know that this is the exact version. So there is a bit of a work that people have to do if they want to go for the simplest option, but it needs some sort of detailed look into which those devices are. So that is another option. I've also seen people exploring using something like a Raspberry Pi using an USB to Ethernet. But there is an issue with that in the sense that if you have connections which are faster and the Pi version 4 is much more powerful, but previous to that is not... I think there is... In the previous version, there is also the sharing of the bus on the hardware front as well. So it may not work for the performance. I've never tried that. I mean, just to let everyone know Raspberry Pi root is something that I've never tried. I think there is a hardware by TP-Link which I think the market it has load balancer router can take like one to four LAN connections and then give you one single LAN. I see there are some... Like I foresee some issues with that hardware, one being it's 100 Mbps limited. So if you are having like a connection more than 100 Mbps, even two connections with 100 Mbps, it will still output you like 100 Mbps stream, right? Because then it does not support. The other problem that I see at the software level is with banking apps or apps which use sticky cookies as Gaurav mentioned. So banking apps will certainly break if the router just does dumb switching of packets, right? Like if you go to netbanking.sdse.com, it will just break in between that say, hey, your session has been locked somewhere else. So unless you have some kind of intelligent splitting, maybe using hostname-based routing, let's say this is hostname. Basically you switch based on host headers, but then this only applies to HTTP. Again, it does not work on other types of connections. Other types of connections may use let's say proxy protocol. But again, as you mentioned, the T-pilling hardware is very common in India, but then having open WRT work on them is difficult. Even I can't run open WRT on my T-pilling router. So one question to you, which router did you end up buying? Is this question to me? Yeah. So I have a, let me, so it's kind of a mini PC that I had picked up in my last trip to Shenzhen directly from the factory. So I just put the link there. But I think it's available on Aliexpress as well as, I think on Amazon as well. So it's a four port kind of a mini PC router, which runs anything that you want it to run. Sorry, I run pfSense on it. And yeah, so that's the one I'm using. Nice. Cool. I'm just posting a link to the device that I saw. It may solve some people's problem if they don't want high bandwidth or can like circumvent net banking not working. So Swapnil, Chinmay on YouTube is asking, unsure if a router, I'm not sure if it's a question though. It looks more like a suggestion. Unsure if it's a router, unsure if a router would work. Most routers have just a single run input, I think. And what you need is a LD switch. Yeah. So the device which I have just put in the chat that that does even in the even in the case of the open WRT one. So let's say the TP link examples, which I was talking about, you could configure, you know, there are certain models where you can actually configure, you know, devices to use instead of a LAN, configure it as a van. So you can have multiple van ports that can that kind of work. Yeah. I think Devang is also mentioning the same using open WRT, we can convert LAN port to it. So it would make like chipset support, right? Wanted like it won't just suffice with software, right? Yeah. So once you load the, once the, once you load the firmware, there are certain models where, you know, in let's say when you load open WRT firmware, it will give you an option of whether you want to configure this as a, you know, as a van or a LAN. So you can select that this can also, this is one additional port that I want it to work as a van port. May not work for the base devices. A lot of base devices or the generic ones which are available here in India, especially may not work for those, but I've seen there are certain models where it does that. Devang has a question. Any suggestion, any suggestion to monitor UDP based graphics for ISP? I didn't quite get the question. Devang, if you can clarify, All right. Yeah. So any, any tools just to monitor UDP based, because sometimes I have observed that websites are working fine, but when it comes to the streaming, okay, real-time traffic, it causing a kind of latency or a jitter or delay in terms of response. So is there any specific tool for UDP or live real-time traffic to monitor? I'm just trying to get an understanding on the question. So one thing, what you could, you could definitely look at is, is use a right pathless probes, if that makes sense. So if you have a probe in your network, so I'm not talking about probes in other networks, but you could use a probe which is in your network and run a measurement using that. And there are measurements for a trace route, if I'm not mistaken, which is can be UDP based as well. So you could use that as a base and look into integrating that into something like Prometheus or using the actual 6-ported, that could be an option. Any other tool per se without, without using, nothing is coming to my mind, but maybe somebody else has any inputs in the meantime, let me try to refresh if something can be, that I can suggest. One thing about the Atlas probes, make sure you start running an Atlas probe which you can maintain, that way you'll start accumulating credits and using those credits, you can start monitoring your network and whatnot and make sure your generation of credits is always higher than your spend. That way you won't ever run out of credits. I think I learned this the hard way. I thought that as soon as I would have an Atlas probe, I'd be able to run all kinds of measurements and turned out that I ran out of credits pretty soon. So yeah. So I am a millionaire. I have millions and millions of credits because of all the probes that I run in a number of networks. So if any of you, I mean my first suggestion to people here in the room as well as people who we watch this later is to host a probe in the network because that's a good thing to do for the community. But in case if you need a credit voucher, feel free to reach out. I can give you a credit voucher which you can use to run measurements because I understand a lot of people do not have compute to install a software probe especially in home networks. So I understand the limitation. Not everyone will have a Raspberry Pi or an Intel NUC or a server like I have in my bedroom right now. So having said that, feel free to reach out if you'd like to run measurements.