Cloudflare’s Post

Not personally, and I guess it stretches the definition of a "server" a bit, but Voyager 1 has to be up there. 48 years and more than 25 billion kms from the nearest sysadmin. Quite incredible. https://science.nasa.gov/mission/voyager/voyager-1/

Multiple 15+ years Solaris and Unix systems, network routers and switches that are approaching two decades and Novell Netware server that has been running since 1997 that is still technically up. Most of these are in protected environments now luckily. I built a system that maintained five nines 99.999% for two years when measured on the application level. There were of course 3 datacenters, 15 servers, multiple redundant datacenter networks with 7 transits and 5 IXP connections with tons of peering and a whole bunch of opensource magic (varnish, nginx, exabgp, powerdns, haproxy...) below the actual service.

16 years, 30 weeks, 4 days, 4 hours, 7 minutes on a network router. There would be concerns about security patches. Addressing that the router was out of production years earlier and keeping running just for demonstration purposes.

Back in my teens I got a hand-me-down 200 mhz laptop that had a burnt out backlight. IIRC I probably had redhat 5.x running on it. It ended up becoming a webserver for my site. Best uptime I had was 843 days before I had to unplug it to move (the battery only held charge for a few mins unfortunately). Maybe not the best uptime out there, but probably one of the more absurd.

When I was in Afghanistan there was an ancient Google Earth server that no one could physically locate. I found it in the back of a shipping container, plugged into an UPS. I want to say it had close to 6 years of uptime. Don't remember the distro it was running, but literally had just been sitting there with nothing happening on it except the daily automated imagery updates.

Ironic timing - our entire platform (donely.ai) has been down for 40+ minutes because a failed Cloudflare D1 export locked our production database with no way to cancel it. Support case #02055784 is open with no response. P1 level request, and we are in the Enterprise Plan!!! The DB is completely inaccessible - no queries, no time-travel restore, nothing works. Would love some of that uptime right about now! 😅

we’ve seen some surprisingly long uptimes in legacy environments, especially where stability was prioritised over frequent patching. It always raises an interesting trade-off between resilience, security posture, and the need for continuous updates.

If your server has been up for literally "years" then it probably hasn't been patched and updated.

13 years+ CentOS server built on consumer grade equipment. Bless that box -- and the APC that supported it thru multiple Florida hurricanes.