24 hours? Wouldn’t it be better to do shorter bouts of scheduled unavailability so unknowing people’s systems will boot up without manual intervention, but still generate lots of nasty logs / alerts?

▲

rollulus 5 days ago | parent | next [-]

I thought the opposite: 24h seems too brief to me, since many of their images are typically for long running servers, some people will receive a painful heads up only next year or later when their K8s pod gets scheduled to a new machine, requiring a (failing) pull.

	▲	alias_neo 5 days ago \| parent [-]
		I'm glad this was top of Hacker News because I hadn't heard about this until now, and we'd only have found out once deployments started failing. It's not always a 5 minute job to switch to a different image with different configuration and retooling required. Fortunately, I started moving us away from Bitnami a little while ago because they started giving me the ick some time back, but a few stragglers remain.

▲

ctippett 5 days ago | parent | prev [-]

If I had to hazard a guess, it's so the downtime is noticed across various different timezones.