Mander

AAAHHHH 😭 Such a relief. This was stressful.

The websites makes use of Contabo's object storage service for storing and serving images. Earlier during the month, I got an e-mail saying that there would be a 12-hour downtime window due to a migration. Alright, no problem.

A bit before this users reported problems with uploads. I figured this was related to the migration.

After the "migration", images still caused problems due to some "administrative rule" preventing uploads. I sent messages daily to Contabo. At first they were responding that "their engineers were looking into it, that this was related to the migration, and that they had no time estimate". OK, I can be patient....

After a few days I lost some of my patience and pressed them more often, until they started sending me generic messages and eventually saying that "the problems were already fixed" and suggesting it may be some configuration error on my side.

This is where it got stressful, because my knowledge of the implementation of the images back-end is superficial. So... That meant I would need to study the details and either fix it myself or show proof that it was their fault.

These past few days I have tried to do some reading and tested configuration changes, but I am the kind of person that puts too much on their plate and I have been quite busy, so this has been an additional stress point.

Today, images suddenly started working (until it came down again due to some logs piling up). This tells me that it was not a configuration issue and it was with high probability something on Contabo's side that they fixed today. Uploads are working as well now. Contabo support never shared any information about the specifics of what went wrong, so I don't know.

This doesn't mean that I don't need to study the details, but at least now the pressure is lower. My plan now is to do some studying on how images are stored and served and once I have a good understanding I can migrate the instance to a dedicated server. I will also look into techniques for making a server more resilient when things do go wrong.