On 12-07-2020 between 14:20 UTC and 16:08 UTC, Atlassian customers using Trello may have experienced service interruptions.

Trello became unavailable due to our database not being able to handle a high level of load coming from our API. This high level of load was triggered by our websockets service hitting a hard limit on the number of network connections allowed, which caused an increase in API traffic due to our clients reconnecting. We provisioned a new database node at twice the size to take over as primary and absorb the load and this resolved the incident.

We are taking a number of actions to prevent these types of incidents in the future, including:

• Strategies for rapidly scaling up our database nodes, as needed in an emergency
• Improved load testing for services
• Additional monitoring/alerting for potential operating system networking limits
• Tooling to assist us with bringing Trello back online more quickly, including improved load shedding capabilities

While we working on those long-term improvements we have put the following immediate short-term measures in place to improve reliability. They are:

• Increased resource provisioning across many parts of our infrastructure
• Increased dashboard monitoring during and post-release, across all clients
• Enhanced capabilities to control which clients can generate traffic and how much

We understand that outages negatively impact your productivity and we apologize for the inconvenience this has caused.