Graphing the ping time between Linode clusters

I needed a test case for my new multiping_async munin plugin so I made a graph of the ping time between Linode clusters:

~~![](<URL url=)http://munin.nix.is/nix.is/v.nix.is/mul … de-day.png">http://munin.nix.is/nix.is/v.nix.is/multipingasynclinode-day.png" />

~~![](<URL url=)http://munin.nix.is/nix.is/v.nix.is/mul … e-week.png">http://munin.nix.is/nix.is/v.nix.is/multipingasynclinode-week.png" />

This is run on a Linode in London, it's interesting how fuzzy Atlanta is compared to the others.

Now all I need is four volunteers each in their own cluster that isn't London to run this so we can graph the ping times from each cluster to everywhere else.~~~~

6 Replies

Cool! I can run it in Newark or Dallas, if you need.

I'd run it in newark, but it looks like funkytastic has that covered.

Well, you guys could either run munin with that script yourselves, or I could come up with some elaborate system where you run the script in cron & save the data. Which I then fetch over HTTP and graph.

I've been running my own solution for this for a while on an internal rig (all five DCs to/from all five DCs) that monitors networking overall. I was let down by how boring the data is, but it is a testament to route stability. It's interesting that when you plot the data out, you can see when routes between datacenters change - latency will abruptly shift from one value to another, and stay rock solid at the new value. Get ready for some surprisingly flat lines, though. :)

Your graphs look good and are damned near a copy of mine…my rig's are a lot flatter with less variance, but I might be sending more packets than you (I do 5 seconds, and I'm using collectd - apples/oranges).

@jed:

Your graphs look good and are damned near a copy of mine…my rig's are a lot flatter with less variance, but I might be sending more packets than you (I do 5 seconds, and I'm using collectd - apples/oranges).

Right, this is just a lazy implementation that runs ping -c 3 every 5 minutes. It's not going to be as good as something proper like collectd or smokeping.

I wrote it to monitor the ping times from the node to users that have shell access on it. I just thought I'd throw a ping for the Linode clusters in there for fun.

@Ævar Arnfjörð Bjarmason:

It's not going to be as good as something proper like collectd or smokeping.
I don't even know if I'd call it 'proper,' really…I went with collectd for better SNMP, which is my system's primary focus. I think they have different uses, and I had a lot easier time setting up Munin than collectd. collectd just gave me a bit more horsepower for certain things. Sucks to have to write your own display UI for the collected RRDs, though, since nothing seems to fit. I think Munin is more of a "apt-get install it and forget it," whereas collectd has required a bit more care and feeding from me to keep going.

I like that new plugins are being developed for Munin, though, and I thought it was quite easy to do so - another Munin strength over collectd. Sorry - didn't mean to drop into your thread and tout collectd, I was just intrigued by your work since I'm a sucker for graphs. :)

I really wonder what the patterns mean sometimes. Occasionally the links will do a sawtooth pattern which is pretty cool-looking, and makes me wonder what network optimization is causing that.

Reply

Please enter an answer
Tips:

You can mention users to notify them: @username

You can use Markdown to format your question. For more examples see the Markdown Cheatsheet.

> I’m a blockquote.

I’m a blockquote.

[I'm a link] (https://www.google.com)

I'm a link

**I am bold** I am bold

*I am italicized* I am italicized

Community Code of Conduct