So we know that’s a pretty strange blog-post title, but it comes from an old Chemistry class long ago, when we were complaining about Engineers and how they like to improve things that are not broken (so that they become broken). Well that’s what we did, and this posting is written in the unlikely hope that this lesson can be passed on.
We were “cleaning up” our HAPROXY server configuration file because, well, we had comment fields in there from prior configuration set-ups and so forth. So what harm can you do when you are “just” removing comment lines from a config file, huh? But hey, while we are there, let’s make things “consistent”. We did. And shortly after, coincidentally, our superbly reliable and brilliant OnlyOffice document server…stopped working.
We thought maybe a router setting had gotten reset in a power outage (we have had that before). So we checked. Nope, not the router. Maybe the DNS was pointing to the inactive backup server? Nope, not that either. What could it possibly be?
We eventually discovered what we did, which is why we now think we are Engineers: we tinkered with something that wasn’t broken, and then we broke it. It was in our HTTPS HAPROXY server configuration setting. This was the ‘new improved’ line (we changed IP addresses in the line below just because we could/should):
server is_onlyoffice 188.8.131.52:443 check
And this was the ‘old, unimproved’ line:
server is_onlyoffice 184.108.40.206:443
We added the ‘check’ as haproxy is able to check to see if a server is alive – but it doesn’t always work with every server, and as we now know, it certainly doesn’t work with ours. So we ‘unimproved’ that configuration line and…we now have a functioning OnlyOffice document server running again, which we also upgraded to the latest version to make sure we hadn’t completely wasted our time at the terminal.
Key takeaway? If it ain’t broke, PLEASE, don’t fix it. Hopefully we can learn that lesson, even if no-one else can.