At approximately 9:50am on Thursday 4/7 an instance was recycled due to a memory limit being reached, causing degraded CRM performance for a subset of workspaces. We scaled these instances up to double the available memory, and scaled this out to add an additional instance for better load balancing, this was applied at 9:55am, and with a short delay as the instances were “warmed up”, regular CRM perfomance was restored by 10:00am.
Following a period of monitoring and reflection, we have re-scaled the instances in, removing the additional instance added, but have continued to keep the instances scaled up to avoid hitting any further memory limits and preventing further performance issues.