Some of guys noticed that we had some issues regarding our webpage a few weeks back. Sometimes the correct page opened and sometimes not. And sometimes you could submit a paper and sometimes ... not.
So, what kind of testers we are if we don’t analyze bugs which happen in production. So - here you are - a “works on my machine” retrospective:
On the 6th of September we decided to finally add SSL support our web page - nordictestingdays.eu.
Considered, done - as it is was straightforward and simple issue to do, request to our hosting company was made and on the same evening we received the confirmation that all is done and configured.
With proper testing attitude, we made initial sanity check to see whether crucial functionality (including submitting a CFP) works. Initially it was fine till at one moment our page snapshot from year 2014 opened up. First thought - well, did somebody just restore something from the backup? Few refreshes and some browsing later, all worked again. Till… some page loading displayed “It works!”. Well, nice that “it works”, but not the way we expect to work it on our machine.
As we knew that we ourselves did not change anything which should result in such behavior, we contacted our hosting provider for the explanation and investigation.
All upcoming events after that can be summarized as following:
There was a two day long constant ping-pong between us and hosting provider. We claimed that at random moments page is not loaded with expected result and hosting provider was unable to reproduce or debug it (though they were pretty certain about the fact that it’s not an actual problem and it might be related to caching, DNS distribution, etc).
It ended up by our team member physically sitting on the line with the helpdesk and letting them to refresh the page till they actually get it reproduced. After success on the reproduction, it still took around a day to get it fixed.
So what actually happened? With the change to SSL, our DNS mapping also changed - new IP address was assigned to our domain. Though, on hosting provider side the IP was mapped to two different servers - so, at some point site was served from one (incorrect) location and at some point it was served from the correct one. Why we saw a snapshot from the older version first? We actually still do not know how come our old site ended up in the initial display and it seems safer not to debug this legacy anymore.
But we are now smarter and already now you all can enjoy browsing our web page safer. We plan to switch all regular http traffic to https also in the foreseeable future, but this time we are much more cautious with this change.