Show HN: Real-time, top-like metrics for Nginx | Hacker News Reader

Show HN: Real-time, top-like metrics for Nginx

github.com

134 pointslebinh33 comments

For those who didn't know about it, please check Apachetop:

https://github.com/JeremyJones/Apachetop

I use it for both Apache and Nginx.

Side question: I saw a project similar to this, but display top in your terminal/browser with a beautiful visual dashboard. I remember seeing it here on HN recently but I can't find it. I mention that because it has a non-ascii visual component. I am interested in looking at the two source codes and see if I could make pretty dashboard out of it...

EDIT:

Thanks for finding the, HNers. I also think it is possible to look at IPythoNoteBook integration.

Fulfills the browser aspect of the question...this?

http://scoutapp.github.io/scout_realtime/

Yes. This one. Thanks. Sorry for hijacking the thread. But trying to gather a list of metric tools I can try.

The only other top I know of is htop http://hisham.hm/htop/ but I don't think that's what you're looking for

could it be go access?

http://goaccess.prosoftcorp.com/

Maybe that one? https://github.com/Unitech/pm2

How about porting it to LUA and embed directly in nginx?

Very cool tool. It seems to be clearing the screen every time it runs though which isn't great, especially when it crashes that hides the backtrace. Consider using ncurses?

On a different note... developers, please use Python 3. Several Linux distributions are now using Py3 as their main python and those scripts with an "env python" hashbang will not work.

Porting your codebase to be python 2+3 compatible is dead simple, and most of the time it is just a matter of using print("foo") instead of print "foo". So please be considerate.

Unfortunately most of the servers I'm working with are still on python2 by default. The latest Ubuntu LTS, 12.04, only has python2 by default so python2 is still my priority. But in this case, a port should be trivia so I'll look into python3 as soon as possible.

Yes, ncurses would be much better but I haven't had time for it yet. Clearing screen is simple enough and work pretty well for me so far so here we are :)

It's a release goal for 14.04 (due in about 30 days) to have no python2 software in the default desktop/server installs. Seems like there's still lots to be done though. https://wiki.ubuntu.com/Python/3

> Several Linux distributions are now using Py3 as their main python

Only Arch Linux does it.

Nice. You can get similar for any HTTP interface via Varnish and Varnishtop, if you're interested, as well.

Does it matter which version of python we use?

Seems like it was made with 2. Will not run on 3.

Looks cool!

Installation instructions would be a nice plus

Thanks, I'll definitely add that, maybe put this on pypi asap. For now you can clone the repo and then run

pip install -r requirements.txt

to install required dependencies.

Doesn't seem to like to read the logs :

./ngxtop.py -l /opt/nginx/logs/access.log ## just pretends like the log is not being written to

For some reason the access log file is defaulted to logs/access.log too~

There was a bug that was overwriting --access-log (-l) option. Should be fixed now in master

This is nice, congrats! Would also like some performance metrics, or at least exec time per script

If what you mean by exec time is request_time, i.e., serving time of a request, then it's pretty close on my agenda as long as you have it in your access log :)

Yes, an average serving time per resource request would definitely be nice :)

Very cool put it on pypi it will be useful with my blogs in aws

I think this is a great idea. Very useful, thank you.

Cool! Why is this not on pypi?

Thanks, I'll put it on pypi soon, just put it on github today

Just a heads up, you're going to want to translate your readme to rst if you want it to render right on pypi.

Am I the only one that bothered when tools like this do not give you some insight on resource consumption/limit?

Even if it's just parsing logs via sqlite, I still want to see the limitations and X per Y usage metrics to see if it can handle monitoring high traffic domain without impacting performance.

You're seriously bothered about the robustness of a 2 hour old project?

Well, if you're announcing a project that monitors a performance critical part of your infrastructure, it seems fairly appropriate to performance test your product.

There is a difference between a project and a product.

The performance impact really depend on your server and traffics. Currently with high traffic server it will take quite some amount of CPU if running for awhile, after 15-20 mins or so, but I'm working on reduce that. So now it mostly useful to know what's going on with your nginx at the moment but not over a long period.

You aren't the only one. I often am bothered by the limitations of my own project. I rarely need outsiders to point out the major issues because I know of so many myself. Some people consider it a talent to blurt out obvious observations though.

Reply on news.ycombinator.com