https://github.com/dustin/statcap

Memcached stat capture/logging tool.
https://github.com/dustin/statcap

Last synced: 3 days ago
JSON representation

Memcached stat capture/logging tool.

• Host: GitHub
• URL: https://github.com/dustin/statcap
• Owner: dustin
• Created: 2012-02-18T21:21:04.000Z (almost 13 years ago)
• Default Branch: master
• Last Pushed: 2012-11-01T07:14:50.000Z (about 12 years ago)
• Last Synced: 2024-11-11T13:06:50.490Z (about 1 month ago)
• Language: Go
• Homepage:
• Size: 145 KB
• Stars: 4
• Watchers: 3
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.markdown

Awesome Lists containing this project

README

        
# Capture memcached Stats

There are millions of these things, and I've written several myself.

This is another one.

The purpose of this tool is to capture stats from a

memcached/membase/couchbase node and store them into a CouchDB for

offline analysis.  This is suitable for unattended operation at a

customer site reporting back home.

# Usage

    Usage of ./statcap:

      -out="http://localhost:5984/stats": http://couch.db/path or a /file/path

      -proto="": Proto document, into which timings stats will be added

      -server="localhost:11211": memcached server to connect to

      -sleep=5: Sleep time between samples

      -stats="timings,kvtimings": stats to fetch beyond toplevel; comma separated

Most of the usage should be obvious, but I'll add even more

description here so people know what's up.

## Out

Where to send the output.  This can be a URL to a CouchDB, or a file

path where a stream of gzipped JSON docs will be written.

CouchDB supports basic auth, so you can do something like this:

    ./statscap -out=http://myuser:[email protected]/secretdb

or a plain file:

    ./statscap -out=file.gz

## Proto

This one probably needs the most explanation, but the rationale is

pretty straightforward:  I want to be able to log additional

properties along with the stat snapshots so I can aggregate the

records more easily.

In order to do this, I supply a JSON document prototype that has a

bunch of properties, and then build the stats on top of it.  For

example:

    {

        "customer": "enterprise rent a bar",

        "mission_date": "2012-02-18T13:30:00",

        "attempt": 382

    }

Those properties will now appear in the final documents.

If your proto document includes properties that match the names of

stats you're capturing, the stats will win.  Don't do that and expect

otherwise.

## Server

The memcached server to talk to.  Binary protocol only and no auth

support currently.

## Sleep

How long (in seconds) to wait between samples.

## Stats

The toplevel stats are always captured and stored as "`all`".

You can specify additional stats to grab here.  The default list

includes some I care about right now and might change in the future.