Running an uptime monitoring program on a VPS?

rockinmusicgv

I want to run an uptime monitoring program on something other than the many raspberry pis I have lying around my room. I know there are many commercially available solutions, but using one of those would make way too much sense.

In any event, my concern is that the program will generate lots and lots of network connections and I don't know how much of an issue that would be. Only a few bytes would be transferred for each connection, so it's not lots of data, just lots of connections being created.

I'm currently running the script on a vm. The load only seems to spike when tmux is having problems redrawing the screen. Otherwise it sits around 1-2%, Averages as I type this are 0.00 0.03 0.00 and it's only using ~256mb of 1g it has access to.

Is this something that can be done on a regular VPS without bothering any neighbors? Am I over thinking the potential problems here?

Decicus

You're likely overthinking the problem, unless you're talking about an extreme number of uptime targets and connecting to them very frequently

rockinmusicgv

What would you say the limit is for this type of thing 1 per minute, 10 per minute, 100 per minute? In my head, 10 requests per minute (~15,000 requests per day) would be enough to cause some sort of problems for someone but I could be wrongI t. It's probably more than what I'd be doing - but if that is acceptable I'd be fine.

yoursunny

VirmAche AUP:
High Network Usage: Customer’s Service cannot have more than 50,000 conntrack sessions at any given time, and cannot use more than the allocated bandwidth.

Each connection creates one conntrack session.
After the TCP connection is closed or the UDP/ICMP flow sees no more traffic, the conntrack session would be released in 5 minutes or less.
Hence, you can open 166 connections per second without violating VirMach AUP.

rockinmusicgv

Right so then even once every half second is well in the range of ok...

rcy026

Unless we are talking hundreds of connections per second you will be ok. Most monitoring solutions use icmp or very basic requests to do their work, the load is most likely totally negligible and not even noticeable when mixed with all other network traffic. The number of connections would probably be the first issue, but again, then we are talking hundred of new connections every second constantly.

chitree

I'd recommend checking out Update Kuma here https://github.com/louislam/uptime-kuma

shallow

Why do you need to check idling servers uptime? If it does not respond to ping then it is power idling.

tetech

@rockinmusicgv said:
Right so then even once every half second is well in the range of ok...

Yes, and you can mux ssh sessions and keep them alive which reduces the number of connections (and latency and bandwidth). For example, in naemon,

        command_line    $USER1$/check_by_ssh \
                -o ControlMaster=auto \
                -o ControlPath=/var/run/nagios/$HOSTNAME$.sock \
                -o ControlPersist=4m \
                -i $USER16$ -H $HOSTADDRESS$ -p $_HOSTSSH_PORT$ -l $USER15$ \
                'check_smtp -H $ARG1$ -p $ARG2$'

But basically you are so far away from hitting the limit that you shouldn't worry.

havoc

Uptime checks are usually >1 min intervals so you'd need a lot of them to piss anyone off

c1vhosting

@rockinmusicgv said:
I want to run an uptime monitoring program on something other than the many raspberry pis I have lying around my room. I know there are many commercially available solutions, but using one of those would make way too much sense.

In any event, my concern is that the program will generate lots and lots of network connections and I don't know how much of an issue that would be. Only a few bytes would be transferred for each connection, so it's not lots of data, just lots of connections being created.

I'm currently running the script on a vm. The load only seems to spike when tmux is having problems redrawing the screen. Otherwise it sits around 1-2%, Averages as I type this are 0.00 0.03 0.00 and it's only using ~256mb of 1g it has access to.

Is this something that can be done on a regular VPS without bothering any neighbors? Am I over thinking the potential problems here?

Uptime Kuma!!

Neoon

The question is always, how long and how much CPU usage.
Usually that is fine even on a cheap NAT VPS for like 3$/y.

Consistent CPU usage is the thing you need to look out for if you don't got dedicated cores.
Second thing, if you are using containers, keep in mind the conntrack limit is shared for every container on the node.

Even on KVM, high amount of connections will result in more load on the Host, that could get you suspended.
If you ask them, they may tell you a limit, like I do sometimes scan the internet and run this on a KVM.

rockinmusicgv

@Neoon said: I do sometimes scan the internet and run this on a KVM.

I feel like this might be the first time someone understands me.

chitree

@c1vhosting said:

Uptime Kuma!!

I 2nd this recommendation