Closed nisdas closed 4 years ago
Imo this is optional. Most of the user slashings we have seen in cluster are due to multiple node instances (ie, different zones) using the same set of the keys. In those cases, having a localize DB will not help.
I think this is a nice to have, but just not sure how much protection it realistically applies...
We haven't tested this in runtime at all.
We can invert the flag so that the default is ON, but I don't think we should remove this flag entirely yet.
I agree with Preston here, it hasn't be thoroughly tested on runtime in the cluster. We should enable it on the cluster for a little bit at least before releasing.
Validators haven't committed any slashable offense right? The network has been stable, less likely for validators to accidentally slash period.
On Wed, Apr 1, 2020, 2:26 PM Preston Van Loon notifications@github.com wrote:
We haven't tested this in runtime at all.
We can invert the flag so that the default is ON, but I don't think we should remove this flag entirely yet.
— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/prysmaticlabs/prysm/issues/5267#issuecomment-607416161, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABPWH5Q7OQCRLIYX6VQZ2VLRKOBKXANCNFSM4LYPY5BA .
Starting this change now.
Currently we do not require our validators to have a db to protect them against accidentally slashing themselves. This should be the default for running our validators in the next testnet restart. Although this might affect our k8s deployments as now all our validator clients will need to have a db. @prestonvanloon would this be possible to implement for our next testnet restart ?