rmoorman / bigwig

like erlang's webtool, but trendy and new

Bigwig is a suite of web-based tools for the Erlang VM, like webtool.

Tools that we currently include custom versions of:

• "etop" - see which processes are using the most cpu/heap/reds/etc
• "rb" - browse SASL reports
• "appmon" - explore application supervision hierarchies

Also included is a VM dashboard with release_handler details, loaded applications, and VM settings.

Conceived for Spawnfest 2011, SMELLS LIKE BEAM SPIRIT.

USAGE

Start up the dev server and head to http://localhost:40829/

./rebar get-deps
./start-dev.sh

or include as a rebar dep and start your own node:

{bigwig, ".*", {git, "git://github.com/beamspirit/bigwig.git", {branch, "master"}}}

rebar get-deps, compile, and then make sure your config includes the following so that binary sasl reports are generated (which are used by bigwig):

{sasl, [
     {errlog_type, error},
     {error_logger_mf_dir, "log/sasl"},      % Log directory
     {error_logger_mf_maxbytes, 10485760},   % 10 MB max file size
     {error_logger_mf_maxfiles, 5}           % 5 files max
   ]}

then run your project:

mkdir -p log/sasl/
./start-myproject-dev.sh
1> bigwig:start().
=INFO REPORT==== 10-Jul-2011::20:59:44 ===
Bigwig listening on http://127.0.0.1:40829/

NOTES

Erlang topology

Run Bigwig as a hidden node, on a distinct VM, and it will connect to nodes of your chosing to monitor/examine. No need to install any applications or releases on nodes you wish to inspect.

Web Architecture

We are using cowboy, with a combination of webservices that return JSON, and websockets to deliver realtime updates to the browser.

JSX Json Library

Modified so term_to_json works with pids, ports, etc. Eg, a pid is converted to {"_type": "pid", "data": "<1.2.3>"} Used when we render complex terms in the browser, like SASL reports.

SASL Report Browsing

'rb' is for reading SASL reports written by the log_mf_h event handler which SASL adds to the error_logger process. The on-disk format/writing looks like it's been around long before disk_log wrapfiles were conceived, since it sort of does what you would nowadays use disk_log for.

The rb module is designed to print text reports to stdout. The bigwig_report_loader module is our alternative to rb which loads the SASL report terms from the file and returns them, which we subsequently convert to JSON and render client side. We also use the SASL-provided renderer code and send the familiar looking log format to the client too.

Would love to install our own handler that writes the reports as documents to couch, or at the very least using disk_log, and with a better separation of loading the terms and rendering them. No time for this yet, and at least with our current approach we don't need people to install anything on nodes they wish to examing using Bigwig.

eTop

A web interface surfacing the information provided by etop with client-side sorting, introspection of pids and modules and the ability to pause and page through results.

AppMon

A 'SpaceTree' browser of a node's application masters and supervision hierarchies. Click nodes to browse the hierarchy. It will automatically update unless live updates are disabled (via the button at the top).

URL schemes, blah

• /vm lists running apps, releases, process count etc

• /rb/reports lists available SASL reports

• /rb/reports?type=error lists SASL reports of a specific type

• /rb/reports/1 returns SASL report id 1

• /rb/stream websocket endpoint for streaming SASL reports as JSON

• /pid/<1.2.3> process_info (POST to send msg to pid)

• /pid/NAME as above, but using registered process name NAME

• /pid/global/NAME as above, but using global registered process name NAME

• /module/NAME NAME:module_info() stuff