2 ms·
Oh please no; give me a hard fail I can localize and fix rather than some kind of awful brownout where various parts of the system just go slow and break things
by NovemberWhiskey 6y ago
Oh please no; give me a hard fail I can localize and fix rather than some kind of awful brownout where various parts of the system just go slow and break things just as badly anyway.
Plus you'd need to be way in the guts of the TLS implementation to achieve this; if you're already there, start generating noise a week ahead of the expiration instead.
Or better, none of the above and automate.
- macintux 6y agoConcur. From working at Basho, one key takeaway with distributed systems is that a hard failure is much easier to remediate than a slow machine. We wanted a database server to fail hard. Running slowly just caused cascading failures. Of course, in this case you're effectively talking about the entire cluster crashing hard, but that's still easier to cope with than every system responding at a snail's pace.
- aluminussoma 6y agoI agree that automation is ideal. But let's face it: most companies haven't. The goal of a business is not to have perfect engineering practices. It is to fulfill customer requests. When there is an outage in the middle of the night, I'd argue that a degraded system buys time to address the issue. Regardless of the mechanism, having a sudden, complete breakage is not ideal for a business.