Over the last couple of months, Postmark’s growth has resulted in some recurring problems for our customers that we wanted to take the time to acknowledge and address.
In order to provide you delivery diagnostic tools like the send activity screen and bounce hooks, we store large volumes of information for all customers’ send activity. This activity and dataset is not used for sending your email, just retrieving the history. As Postmark usage has grown, the storage of transactional messages had grown to a point where customers were encountering timeouts on activity screens, and delays in the processing of bounced emails. It became clear that we needed to evaluate our options to eliminate these problems.
A good example is today’s activity page issues. Our current servers were responding very slowly in the morning which caused delays in updating the In queue status to Sent. The emails were sent immediately, but updating those records in our database were delayed. To improve response time we had to trim some of the older records in our database, which is current set to the last nine days. We managed to clean up some records and improve the speed, but at the same time a small number of the In queue records were deleted by accident. Just to be clear, all emails were sent, but since the In queue records were removed, the history for those records are no longer there.
We’re in the process of installing new hardware and some new methods to store and retrieve this data. The goal is to not only improve the response time of the activity pages, but also improve our internal response and queue times for sending messages as we grow. We would also like to start storing at least 30 days of send activity. We’re confident that these changes will have a significant impact on the way you use Postmark and the way we manage it.
We’ve been working hard to get search times faster & store emails for longer periods of time, independent of the growth of Postmark and its customers. The migration is taking longer than we’d hoped, but we’re thankful to be in the final stages of testing. I plan to do a follow-up after the migration is complete and explain the decisions we made as well as the software and hardware that is running it.
Most of all, we’re sorry for the pain this has caused you. We can’t thank our customers enough for being patient and understanding while we complete our migration. You can look forward to an announcement very soon about the improved performance and service enhancements.
This post was originally published May 06, 2011