I was actually referring to my ongoing situation here. A site went down for ßome time but I didn't notice it. And by the time I setup Uptime Kuma & ntfy, it came to my attention but slightly too late.
Being one of those popular PHP services, I thought we were hacked.
The log would show several attempts to access various folders by hackers. And I could not visit the index page. Other pages would show gibberish contents.
Past lessons had proved the need to have ZFS snapshots on the host and remote locations. Rolling back snapshots restored the site. Notwithstanding, there were still some awkwardness in the service. I have learnt over time that Unix, FreeBSD in this case, has one and only different approach to running a service.
One of such awkwardness is that the site becomes inaccessible - CDN throwa error - as soon as the config file of the web service changes without reloading the server. I checked online if nginx automatically applies changes despite being sure that it does not.
In summary, ZFS rollback saved the team you mentioned like it did for us. Another lesson learnt is to set up a crom job to automatically upgrade the service/app DB without interaction should it's version be unconsciously upgraded.
#freebsd #unix