Skip to main content

Scaling pains changing ldap to mysql was like changing engine of a nascar during race

When you have a big customer base then you get lots of interesting problems like :
  1. Customers trying to store 400K files in one folder
  2. Customers trying to dump 100+ TB of files in one account
  3. Customers with more than half of their data in Trash
  4. Customers wanting to create> 64K users in an account
  5. Customers with 25K+ users trying to do search/load/sort on users listing
Our startup used to use LDAP to store users and customer metadata and Mysql to store files/folder metadata. 

Lately we had been hit by #4 and #5 item and that is causing scaling issues because no matter what we do ldap write performance sucks when we go beyond some million users.  Earlier ldap used to scale because we had 30-40 ldaps but to reduce ops management issues we consolidated then to 4 per DC and it worked initially but lately with #4 and #5 its not scaling fine.  Ldap is an alien to most programmers, there are only 2-3 guys in the team that knows a little bit about it so most of the times it remains orphan and being involved in every prod issue regarding ldap is not fun.

So we started on a project to replace ldap with mysql but problem is that ldap was in the company from day one so replacing it isnt that easy.  We use SOA and there are many moving parts that uses ldap, converting ldap to mysql required upgrading all services in all data centers at same time which sounds both risky and a hard sell to ops/management team.  So how do you change an engine of a running race car?

Well thats where interface based programming and feature flags comes into picture.

We have all our code related to ldap in one layer implementing DirectoryService Interface



To move to mysql we introduced two fields on each customer ldap_url and mysql_url and prepopulated ldap_url for all customers. We then created RoutingDirectoryService that for each request will determine whether customer is on ldap or mysql and routes the call accordingly to LdapDirectoryService or SqlDirectoryService.

We had to follow the same pattern in all other services and some were written in python so we had to duplicate the effort.

But luckily we were able to achieve the objective with less disruption to release process and we released the code in sleeper mode to prod one service at a time. The code was in sleeper mode because no customers were moved to mysql until we upgraded all services to use router logic.

Last night we migrated 100 customers to mysql and so far we saw only one issue.  We would continue to migrate more customer over the month  and eventually when ldap is gone we would remove RoutingDirectoryService.

4 years ago if we had done this then it would have been easy as no of customers were less but now we have so many customers that  we need to go feature flag way. Also customers rely on us 24X7 so we cant take risk of changing the engines of all car in a running race at the same time so we are now doing one car at a time :) and putting them back on race track.

Comments

Popular posts from this blog

RabbitMQ java clients for beginners

Here is a sample of a consumer and producer example for RabbitMQ. The steps are
Download ErlangDownload Rabbit MQ ServerDownload Rabbit MQ Java client jarsCompile and run the below two class and you are done.
This sample create a Durable Exchange, Queue and a Message. You will have to start the consumer first before you start the for the first time.

For more information on AMQP, Exchanges, Queues, read this excellent tutorial
http://blogs.digitar.com/jjww/2009/01/rabbits-and-warrens/

+++++++++++++++++RabbitMQProducer.java+++++++++++++++++++++++++++
import com.rabbitmq.client.Connection; import com.rabbitmq.client.Channel; import com.rabbitmq.client.*; public class RabbitMQProducer { public static void main(String []args) throws Exception { ConnectionFactory factory = new ConnectionFactory(); factory.setUsername("guest"); factory.setPassword("guest"); factory.setVirtualHost("/"); factory.setHost("127.0.0.1"); factory.setPort(5672); Conne…

Logging to Graphite monitoring tool from java

We use Graphite as a tool for monitoring some stats and watch trends. A requirement is to monitor impact of new releases as build is deployed to app nodes to see if things like
1) Has the memcache usage increased.
2) Has the no of Java exceptions went up.
3) Is the app using more tomcat threads.
Here is a screenshot

We changed the installer to log a deploy event when a new build is deployed. I wrote a simple spring bean to log graphite events using java. Logging to graphite is easy, all you need to do is open a socket and send lines of events.
import org.slf4j.Logger;import org.slf4j.LoggerFactory; import java.io.OutputStreamWriter; import java.io.Writer; import java.net.Socket; import java.util.HashMap; import java.util.Map; public class GraphiteLogger { private static final Logger logger = LoggerFactory.getLogger(GraphiteLogger.class); private String graphiteHost; private int graphitePort; public String getGraphiteHost() { return graphiteHost; } public void setGraphite…

What a rocky start to labor day weekend

Woke up by earthquake at 7:00 AM in morning and then couldn't get to sleep. I took a bath, made my tea and started checking emails and saw that after last night deployment three storage node out of 100s of nodes were running into Full GC. What was special about the 3 nodes was that each one was in a different Data centre but it was named same app02.  This got me curious I asked the node to be taken out of rotation and take a heap dump.  Yesterday night a new release has happened and I had upgraded spymemcached library version as new relic now natively supports instrumentation on it so it was a suspect. And the hunch was a bullseye, the heap dump clearly showed it taking 1.3G and full GCs were taking 6 sec but not claiming anything.



I have a quartz job in each jvm that takes a thread dump every 5 minutes and saves last 300 of them, checking few of them quickly showed a common thread among all 3 data centres. It seems there was a long running job that was trying to replicate pending…