Weeks 05/28/2017 - 06/11/2017 Assessing HikariCP and MariaDB driver

Trying to back fill a few weeks of tech reports before I forgot what I did:

Spend 2 weeks assessing 2 libraries: 
 - HikariCP: https://github.com/brettwooldridge/HikariCP
 - MariaDB Connector/J: https://github.com/brettwooldridge/HikariCP

Context:
Moving our database from mysql to AWS aurora and MariaDB is the recommended driver library. HikariCP is the most popular jdbc connection pool library and I'm thinking about upgrading from BoneCP to HikariCP.

HikariCP:
Upgrade was fairly simple and configuration can be easily translated to the matching configuration in BoneCP. 
The major blocker for me is performance: tested HikariCP integration production environment and compared result to BoneCP, queries using HikariCP is consistently slower than queries using BoneCP. A frequently used simple select (for auth) takes on avg 1ms with BoneCP, but 2ms on HikariCP, another frequently used update query takes 2ms with BoneCP but 3mc with HikariCP.
More details:
http://xingdu.blogspot.com/2017/05/performance-between-bonecp-and-hikaricp.html
https://github.com/brettwooldridge/HikariCP/issues/900

MariaDB driver:
Assessed 2.0.1 version and ran into compatibility issue with Aurora. Spent a day debugging and found out it was due to a flag introduced in 1.6.0 / 2.0.0 (usePipelineAuth defaults to true and apparently doesn't work with Aurora yet)
Filed a ticket with developer:
https://jira.mariadb.org/projects/CONJ/issues/CONJ-477
And ran into another exception during failover testing:
https://jira.mariadb.org/projects/CONJ/issues/CONJ-484
1st one is easily resolved by downgrading to 1.5.9 version, but 2nd one has not been fixed yet, which blocked me from moving forward.

With the help of tcpdump, I can basically tell how mariadb driver works for Aurora failover: it queries a table in aurora which records the instance end-points behind a cluster end-point as well as the role, and the driver tries to be smart about which one to connect to during failover.
I'm personally not a big fan of this idea. To me the client (in this case driver) should be as dump as possible. If connection is disposed per query, and every time we create a new connection with the cluster endpoint, we'd be fine since the cluster endpoint will always point to the valid instance, even during failover(excluding DNS cache). So my call on that is not to use MariaDB driver and continue with the stable mysql driver we had before.

By setting connection age to be 1 min, the possible window for data loss is up to 1 min. And on top of that I do have a my own ip change detector running per 15s, meaning during failover it will catch up the change in less than 15s. This is already better than what MariaDB can offer and I'm happy with staying with the same stable libraries.

Comments

Post a Comment

Popular posts from this blog

Week 07/02/2017 - 07 /09/2017 Rewrite Memcached Cluster Accessor

Spent the week re-writing memcached accessor & configuration for our java server to support clustering. Initiative of this project: Multiple memcached clusters are used in my project and some of them need to be accessed by multiple services written in different languages, one of which being PHP. The old accessor does not support clustering at all, not even to mention ketama consistent hashing (on the roadmap). While other services do utilize clustering for memcached, the java server accessor create one memcached client per host in the cluster and run a multi-write &/ multi-read. On top of that manual hashing was done to decide which cluster a particular key goes to (based on an additional parameter, for instance, user id) Old format: memcached.name=<cluster_name> memcached.hosts.0=<mc00, mc01> memcached.hosts.1=<mc02, mc03> To fully utilize clustering support from memcahed library, a new format was proposed: New format: accessor.<accessor_name...
Read more

Week 10/22/2017 - 10/29/2017 Repository migration & Deployment process improvement

  In the past week I've been working on mainly one thing: migrating our existing repositories from bitbucket to github.   Git is the first source control tool I used and has become dominant over the past few years. To be honest I'm surprised to see my project uses mercurial.    The task of migration is fairly simple: fast-export will do the heavy lifting for you. Here 's an article on how to do it. The reason why it took me a lot of time is: my new project has baked a lot of secrets / keys inside source code and now it's going to be more public than only accessible by my group.   After I started digging into these secrets, I've discovered more issues:  - configuration management and secret management are done with the same configuration file  - configuration and source code is not a perfect 1:1 map. With a lot of copy-paste, many sections of configuration are not even being used in code  - deployment is a mess. some repos have dedicated (...
Read more

Week 07/09/2017 - 07/16/2017 Profiling & Performance Improvement

Image
My primary focus during the past week is improve the performance of event handling system on the java server. Context: We use one thread for processing room transaction related user actions. For instance, user move from one room to another. I added instrumentation for profiling data a few weeks ago and noticed that user disconnect is taking a long time to process while other transactions on the thread is usually done with 5ms. I added timer for sections of code that processes such even and ran profiling on production this week, and the result is astonishing. The majority of the time was spent on a synchronized block. In short, we kick off a call to update database on a different thread and move on with the current thread, and the next thing current thread does is to synchronize on user object. Unfortunately the database updater job locks the user object as well so essentially we're doing the database update on the same thread. And the worst thing is, the lock in database upd...
Read more