Managing the logs of your (Rails) applications - RailsWayCon 2011

download Managing the logs of your (Rails) applications - RailsWayCon 2011

If you can't read please download the document

Transcript of Managing the logs of your (Rails) applications - RailsWayCon 2011

Managing the logs of your (Rails) applications

Lennart Koopmann, RailsWayCon 2011

www.lennartkoopmann.net / @_lennart

About me

23 years oldLiving in HamburgRails developer at XING AGDeveloper of Graylog2

What is this log management stuff?

Even grepping over flat files can be log management.

Log Management Maturity Scale

Log management has different levels Raffael Marty set up a scale for that.

Level 0

Do not collect logs at all.

Level 1

Collect logs. Mostly simple log files from email or HTTP servers.

Level 2

Use the logs for forensics and troubleshooting. Why was that email not sent out? Why was that HTTP 500 thrown?

Level 3

Save searches. The most basic case would be to save a grep command you used.

Level 4

Share searches. Store that search command somewhere so co-workers can find and use it to solve the same problem.

Level 5

Reporting.

Level 6

Alerting. Automate some of your troubleshooting tasks. Be warned automatically instead of waiting for a user to complain.

Level 7

Collect more logs! We may need more sources for some use cases Like multi-line application logs, firewall logs or even physical access logs.

Level 8

Correlation. Manual analysis of all this new data may take too long Correlate different sources.

Level 9

Visual analysis.

Next levels

Pattern detection, interactive visualization, dynamic queries, anomaly detection, more sharing.

Collecting logs

Two different types.

Type 1

Logs automatically generated from a service. For example apache2.log or mail.log Usually huge amount of structured, but raw data.

jira.graylog2.org:80 x.x.x.x - - [29/May/2011:01:47:38 +0200] "GET /browse/WEBINTERFACE-21?page=com.atlassian.jira.plugin.system.issuetabpanels%3Aall-tabpanel HTTP/1.1" 200 7639 "-" "Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)"

Type 2

Logs sent directly from within your application. Triggered for example by a log.error() call or an Exception catcher. - Possible to send structured via for example GELF

2011-05-29 18:55:51 +0200 [payment] Could not validate credit card: Got HTTP 404 from example.org

How to send your logs

Don't store the logs in flat files. Send them somewhere to get more value out of them.

Syslog

Syslog adapters for Rails are available and work pretty good.

GELF

Graylog extended log format Let's you structure your logs. Also check out structured syslog.

Ruby library, Rack exception notifier and Ruby logger available. (www.graylog2.org)

{ 'message':'[pay] ZOMG credit card invalid', 'full_message':'Stacktrace.\nSome env vars','host':'www19', 'file':'/var/www/app.rb','line':2638, 'level':1, '_something':'foo', '_something_else':'bar' }

AMQP

Guaranteed and ordered delivery. Very flexible. Easily subscribe to the flow. Use routing keys to structure origin of the logs. Hell yeah, use this if you have an AMQP bus available. (or build one)

Check out https://github.com/paukul/amqp_logging

Throw the messages out of your app like a hot potato

Loose coupling! Your logs should always leave the application without interfering it! Prefer UDP over TCP, decouple AMQP log transports. Catch all exceptions and get back into the app flow.

Add more value to your logs

For example pre-generate geo information for IP addresses or integrate the time_bandits gem.

https://github.com/skaes/time_bandits

Completed in 680.378ms (View: 28.488, DB: 5.111(2,0), MC: 5.382(6r,0m), GC: 120.100(1), HP: 0(2000000,546468,18682541,934967)) | 200 OK [http://127.0.0.1/jobs/info]

Can generate a deep insight view of your application performance when used with LogJam: https://github.com/alpinegizmo/logjam

Where to send your logs

There are a lot of tools available.

Hosted services: Loggly

www.loggly.comDynamic pricing based on your usageFree for 200MB/day with 1 week retention timeUDP/TCP/HTTP API as input for syslog

Hosted services: Splunk

www.splunk.comTwo license types: Free / EnterpriseSupports any raw input

Two more hosted services:

www.papertrailapp.comwww.logentries.com

Open source tools: Logstash

www.logstash.netCollect, parse and store logs for later useInput Filter OutputPlays very well with Graylog2

Logstash inputs

For example: AMQP, file, redis, stdin, syslog, tcp, stomp, twitter

Logstash filters

For example: date, field, grep, grok, multiline

Logstash outputs

For example: amqp, elasticsearch, gelf, mongodb, redis, stdout, tcp, websocket

Open source tools: Graylog2

www.graylog2.orgAccepts syslog (TCP/UDP) and GELF (+ AMQP)Rails web interface for filtering, analytics, alerting, reporting, Stores in MongoDB

Log management use case: API consumer monitoring Something different from the usual alerting, monitoring and reporting.

Pre-processor script (or Logstash) parses raw access log (possibly via AMQP), combines multi line log messages of API engine and extracts value.

oauth_consumer_key, severity, http_status_code, processed (controller#action)

Pre-processor sends the extracted value including the raw message to Graylog2.

n.notify { :severity => 4, :short_message => UsersController#show [500], :full_message => full_msg, :_oauth_consumer_key => foo, :_processed => UsersController#show, :_http_status_code => 500, ... ...}

Now use Graylog2 and the MongoDB shell to answer questions like...

What consumers are still using the deprecated find user by email call?

What errors are caused by the iPhone application?

Which applications keep causing errors?

Which consumers are inactive?

How many calls are done by the iPhone application and how many were it a month ago?

Extract everything you might need from the message in a structured format you can easily parse and query later.

You already have all the data you need!

Q & A

@_lennartwww.lennartkoopmann.net