1173 views • Outage Dissections
An edge case took down GitHub 🤯
GitHub experienced an outage where their MySQL database went into a degraded state. Upon investigation, it was found out that the outage happened because of an edge case. So, how can an edge case take down a database?
The outage happened because of an edge case which lead to the generation of an inefficient SQL query that was executed very frequently on the database. The database was thus put under a massive load which eventually made it crash leading to an outage.
Automatic retries always help in recovering from a transient issue. During this outage, retries made things worse. Automatic retries added the load on the database that was already under stress.
Now, we take a look at a fictional example where an edge case could potentially take down a DB.
Say, we have an API that returns the number of commits made by a user in the last
n days. The way, this API could be implemented is to get the
start_date as an integer through the query parameter, and the API server could then fire a SQL query like
SELECT count(id) FROM commits WHERE user_id = 123 AND start_time > start_time
In order to fire the query, we convert the string
start_time to an integer, create the query, and then fire it. In the regular case, we get the correct input and then compute the number of commits and respond.
But as an edge case, what if we do not get the query parameter or we get a non-integer value; then depending on the language at hand we may actually use the default integer value like
0 as our
There is a very high chance of this happening when we are using Golang which uses
0 as the default integer value. In such a case, the query that gets executed would be
SELECT count(id) FROM commits WHERE user_id = 123 AND start_time > 0
The above query when executed iterates through all the rows of the table for a particular user, instead of the rows for the last 7 days; making it super inefficient and expensive. The above query would put a huge load on the database and a frequent invocation can actually take down the entire database.
LIMIT 1000would have made you iterate over 1000 rows in the worst case.
If you like what you read subscribe you can always subscribe to my newsletter and get the post delivered straight to your inbox. I write essays on various engineering topics and share it through my weekly newsletter.
444 views • 31 likes • 2022-07-06
Outages are inevitable; but we should design our architecture such that if a component is down, it should not lead to a ...
1059 views • 58 likes • 2022-07-01
Distributed Systems are prone to problems that seem very obscure. GitHub had an outage because a set of nodes in the Zoo...
1165 views • 81 likes • 2022-06-29
So, how are databases managed in production? When the master goes down, how a replica is chosen and promoted to be the n...
A set of courses designed to make you a better engineer and excel at your career; no-fluff, pure engineering.
Being a passionate engineer, I love to talk about a wide range of topics, but these are my personal favourites.
Arpit's Newsletter read by 17000+ engineers
🔥 Thrice a week, in your inbox, an essay about system design, distributed systems, microservices, programming languages internals, or a deep dive on some super-clever algorithm, or just a few tips on building highly scalable distributed systems.
Powered by this tech stack.