Dear colleagues, As many of you have noticed, we have been experiencing some technical issues with RIPE Atlas over the past few weeks. I would like to explain some of these in this email. Earlier this year, we began the process of migrating our infrastructure from our on-premise and co-located data centres to a cloud-based solution. Significant work has gone into rewriting code for multiple components and to renew the underlying infrastructure, to containerise and adapt it for the new cloud environment. In the process of this, we have encountered unforeseen bugs and bottlenecks. The RIPE Atlas team has been working to fix these issues as we find them. We are largely done with the migration of the controlling infrastructure so we hope to have a more stable system in the coming days. We are deploying new code and fixes fairly often to prevent these issues from arising, but it is possible that you might experience unexpected disconnects, but hopefully on a smaller scale than the past weeks. In the coming weeks, we will also begin the migration of the website (the user interface and the API). To clarify some questions that we heard on the list: - We will be looking for a solution to compensate credits for the missing connection times. - In the vast majority of cases, we don’t expect data loss when probes are not connected. The measurement results are delayed until the probes can reconnect and submit the results. - We apologise for the many connection and disconnection emails you might have received when your probe is disconnected. Please note that you can change your email notification preferences on your probe status page. I understand that many of you will want a more detailed explanation. We will be publishing an article on RIPE Labs, and we plan to present our work at the MAT Working Group session at RIPE 89. Regards, Robert Kisteleki RIPE NCC