4 ms·
Less likely since it looks fine from GCP status page. Hmm, it seems that Twitter already figured it out, configuration change issues again.
by azurezyq 7y ago
Less likely since it looks fine from GCP status page.
Hmm, it seems that Twitter already figured it out, configuration change issues again.
- marenkay 7y agoProbably because we all use Kubernetes and YAML files and 100% of configuration failures are "oh shit, I used two spaces instead of 4, we're fucked".
- Jach 7y agoSomething like this is my bet too, there was a recent post somewhere called something like "why all outages are due to a configuration change". There are monocultures in site reliability ops for big companies, "configuration over code" but with heavy automation too. From my outside view it seems there's a tradeoff when you do that between more frequent smaller issues and less frequent bigger issues. Also reminds me of Google's move away from eventual consistency because with their infrastructure they can make a CP system highly available in practice... except when it isn't, due to a botched configuration change.