Blue Canvas Outage History

A fix has been implemented and we are monitoring the results.

We have identified the cause of the intermittent delays affecting branch synchronization and related processing over the past day. A fix has been deployed and the affected processing infrastructure has been replaced. All synchronization services are operating normally and we are monitoring closely to confirm the fix holds under full load. Synchronization jobs that were delayed during this period have resumed automatically — no action is required from customers, and no data was lost.

The "ring" that serves syncs and deploys is stable again. This should get the syncs and deploys running as usual. In the meantime we'll keep an eye out and act immediately if something changes.

Earlier today we identified an issue where ~9% of branches experienced extra delay on both syncs and deploys. The fixes from yesterday helped containing the issue so the other 91% of branches could continue as normal. We've resolved this issue by scaling up the units that have to do the work. There are currently no known issues, but we are looking for a sustainable solution, which may involve one or more releases. In technical terms: The scaling and self-repair configuration we've used for the last year is no long viable. We've greatly optimized on CPU, to the point where it is no longer a reliable metric for scaling in or out. We've disabled the scaling, and manually put the system on a configuration that works right now, and continue to fine tune it while we work on a permanent and automated solution.

We identified the root cause. One of our queued jobs, scheduled to run every 10 seconds, began taking several minutes to complete. These long-running jobs accumulated and consumed all available worker capacity. We've deployed two fixes: - Concurrency limit, only one instance of this job can run at a time, preventing pile-up. - Rewritten algorithm, the job now resolves in ~1 second instead of minutes. We're continuing to monitor.

Due to a recent, Salesforce-mandated security upgrade, all branches need to be reconnected.

The queues seem to be cleared up and everything is working as expected. We've had to clear some of the queues. We have not yet identified the exact job that was delaying the system. However we've found some great optimizations and working on a long term fix for those meanwhile we are monitoring the situation. If your deployment is still showing as "Nothing to merge from" please use "Discard results and validate again" in the top right dropdown of the deployment request view.