Daniel Smith
Daniel Smith
cc @jpbetz, whom I've tried to sell on this in the past :)
Looks like Joe previously proposed this in #10893, but that went stale :(
That sounds great to me! How do I do that?
re: a): I do not think it is a good idea to do this in the storage layer, since it has a bunch of stuff that is irrelevant to the...
OK I commented there, if that's sufficient we can close this.
Automated testing is good, but the problem is not just etcd bugs that are outside of our hypothesis space (which already testing is iffy at finding) -- it's actually not...
I've thought some more about this and I think significant gains can be had just from storing a hash of each {key,value} pair, so that corruption of individual values can...
The benefit of hashing every KV pair: * you can check the hash when the KV is an *input* to a transaction * on finding a corrupt key, you can...
only one minor comment, probably ready to go otherwise?