Gaia-6002 Postmortem

The last testnet, gaia-6002, halted about a week and a half ago. The development team has since isolated root causes, debugged, and pushed fixes for all discovered issues (see the tracker and individual explanations on each issue).

These bugs were primarily caused by several subtle edge cases in the staking/slashing logic, which tends to be complex since it must implement Cosmos’s proof-of-stake algorithm in an efficient manner for an unbounded number of validator candidates, delegators per validator, and validators per delegate. We are taking multiple steps to avoid such bugs in the future: simplifying the validator update logic, writing a randomized simulation testsuite, and experimenting with new methods of collaborative code review.

Meticulous user-submitted bug reports have been instrumental in our debugging. Please continue to submit issues to the Cosmos SDK repository for possible state machine errors and UX flaws alike, and make sure to include the steps to replicate your situation or relevant transactions / block heights for actions taken on a live testnet.

Cheers - see you on the new testnet, gaia-7000, launching soon!

Thanks just lets wait for gaia-7000

Excellent! Can’t wait for gaia-7000!

Great update. Thank you. Look forward to the next testnet!

Issue: when (pex=false), validator losses peers and its not able to reconnected to this lost peer

original issue: cosmos/cosmos-sdk#1458 opened by @iammelea

Just for my understands.
This issue is open and now in Tendermint; so, is not resolved yet. for the gaia-7000 for example.

And is related to the update from Terdermint and not for Cosmos.
I am ok?

Because if i am ok.
Validator need to set
Pex=true
In gaia-7000.
And waith for the update from Tendermint for can uses
Pex=false
For work with sentrys, and don’t get off-line

Just that.