Cloudflare Outage Revisited: Cause & Effects

Image: Mashable

Users woke up on Tuesday to the news of major websites going down due to the Cloudflare outage. While the issue has been fixed, the effects and causes need to be known. The details released on the company’s blog revealed that the outage affected traffic in 19 of their centers. And the 19 locations happen to handle a majority of their global traffic.

Cloudflare reported that the outage was caused by a change that happened to be part of a long-running project to raise the flexibility in the busiest locations. Cloudflare changed the network configuration in those locations, leading to the outage at exactly 06:27 UTC. However, at 06:58 UTC, the first data center came back online; by 07:42 UTC, all data centers were online and working correctly.

The new network configuration has been a project the company has been working on over the last 18 months. The aim, as earlier said, was to convert all its business locations to a more flexible and resilient architecture. Within the time, the company was able to internally convert 19 of its data centers to the Multi-Colo PoP (MCP) network. The data center locations in alphabetical order include Amsterdam, Ashburn, Atlanta, Chicago, Frankfurt, London, Los Angeles, Madrid, Manchester, Miami, Milan, Mumbai, Newark, Osaka, San Jose, São Paulo, Singapore, Sydney and Tokyo.

The outages depend on the location of users in the world. Your location would not allow you to access web services and websites that relied on Cloudflare. However, some locations did not experience the outage, and Cloudflare worked normally. The company took responsibility for the error and apologized to users. They further reiterated that the error was not an attack or malicious activity.

Speaking further on the new architecture, the company said it is designed as a Clos network which happens to be an added layer of routing. The new layer creates a mesh of connections that allows Cloudflare to seamlessly disable and enable parts of the internal network in a data center for maintenance or to attend to an issue.

Cloudflare further revealed that the new architecture had provided them with significant reliability improvements. It has also enabled them to run maintenance in those locations without disrupting customer traffic. The 19 locations recall carrying heavy traffic and any problem within those locations wild range of impact as experienced on Tuesday morning.

According to Downdetector, the list of major websites that were taken down by Cloudflare’s downtime include Twitter, Amazon Web Services (AWS), Discord, Zerodha, Shopify, and Canva. Sites like Udemy, Splunk, Quora, Crunchyroll and crypto exchanges like WazirX, Coinbase, FTX, Bitfinex and OKX also got hit. However, all these sites were later accessible after the fix by Cloudflare.

The 19 locations in question are only about 4% of the entire network of Cloudflare; the outage impacted 50% of total requests. The most important part of the occurrence was the proactive nature of Cloudflare in trying to reduce the event's impact.

Source

#Tech
#Cloudflare