Hive: How we strived for a clean fork

The DAO soft-fork try was troublesome. Not solely did it prove that we underestimated the unwanted side effects on the consensus protocol (i.e. DoS vulnerability), however we additionally managed to introduce a knowledge race into the rushed implementation that was a ticking time bomb. It was not very best, and although averted on the final occasion, the quick approaching hard-fork deadline regarded eerily bleak to say the least. We wanted a brand new technique…

The stepping stone in direction of this was an concept borrowed from Google (courtesy of Nick Johnson): writing up an in depth postmortem of the occasion, aiming to evaluate the foundation causes of the problem, focusing solely on the technical points and applicable measures to forestall recurrence.

Technical options scale and persist; blaming folks doesn’t. ~ Nick

From the postmortem, one attention-grabbing discovery from the attitude of this weblog put up was made. The soft-fork code inside [go-ethereum](https://github.com/ethereum/go-ethereum) appeared strong from all views: a) it was completely lined by unit assessments with a 3:1 test-to-code ratio; b) it was completely reviewed by six basis builders; and c) it was even manually dwell examined on a personal community… But nonetheless, a deadly information race remained, which might have doubtlessly precipitated extreme community disruption.

It transpired that the flaw might solely ever happen in a community consisting of a number of nodes, a number of miners and a number of blocks being minted concurrently. Even when all of these situations held true, there was solely a slight likelihood for the bug to floor. Unit assessments can’t catch it, code reviewers could or could not catch it, and handbook testing catching it could be unlikely. Our conclusion was that the event groups wanted extra instruments to carry out reproducible assessments that will cowl the intricate interaction of a number of nodes in a concurrent networked state of affairs. With out such a device, manually checking the assorted edge instances is unwieldy; and with out doing these checks repeatedly as a part of the event workflow, uncommon errors would turn out to be unimaginable to find in time.

And thus, hive was born…

What’s hive?

Ethereum grew massive to the purpose the place testing implementations grew to become an enormous burden. Unit assessments are advantageous for checking varied implementation quirks, however validating {that a} consumer conforms to some baseline high quality, or validating that purchasers can play properly collectively in a multi consumer surroundings, is all however easy.

Hive is supposed to function an simply expandable take a look at harness the place anybody can add assessments (be these easy validations or community simulations) in any programming language that they’re snug with, and hive ought to concurrently be capable of run these assessments towards all potential purchasers. As such, the harness is supposed to do black field testing the place no consumer particular inner particulars/state may be examined and/or inspected, quite emphasis can be placed on adherence to official specs or behaviors below totally different circumstances.

Most significantly, hive was designed from the bottom as much as run as a part of any purchasers’ CI workflow!

How does hive work?

Hive’s physique and soul is [docker](https://www.docker.com/). Each consumer implementation is a docker picture; each validation suite is a docker picture; and each community simulation is a docker picture. Hive itself is an all encompassing docker picture. It is a very highly effective abstraction…

Since Ethereum clients are docker photographs in hive, builders of the purchasers can assemble the very best surroundings for his or her purchasers to run in (dependency, tooling and configuration sensible). Hive will spin up as many situations as wanted, all of them operating in their very own Linux methods.

Equally, as test suites validating Ethereum purchasers are docker photographs, the author of the assessments can use any programing surroundings he’s most aware of. Hive will guarantee a consumer is operating when it begins the tester, which may then validate if the actual consumer conforms to some desired habits.

Lastly, network simulations are but once more outlined by docker photographs, however in comparison with easy assessments, simulators not solely execute code towards a operating consumer, however can really begin and terminate purchasers at will. These purchasers run in the identical digital community and might freely (or as dictated by the simulator container) join to one another, forming an on-demand personal Ethereum community.

How did hive assist the fork?

Hive is neither a substitute for unit testing nor for thorough reviewing. All present employed practices are important to get a clear implementation of any characteristic. Hive can present validation past what’s possible from a median developer’s perspective: operating in depth assessments that may require complicated execution environments; and checking networking nook instances that may take hours to arrange.

Within the case of the DAO hard-fork, past all of the consensus and unit assessments, we wanted to make sure most significantly that nodes partition cleanly into two subsets on the networking degree: one supporting and one opposing the fork. This was important because it’s unimaginable to foretell what hostile results operating two competing chains in a single community might need, particularly from the minority’s perspective.

As such we have carried out three particular community simulations in hive:

• The first to test that miners operating the total Ethash DAGs generate right block extra-data fields for each pro-forkers and no-forkers, even when attempting to naively spoof.

• The second to confirm {that a} community consisting of combined pro-fork and no-fork nodes/miners accurately splits into two when the fork block arrives, additionally sustaining the break up afterwards.

• The third to test that given an already forked community, newly becoming a member of nodes can sync, quick sync and light-weight sync to the chain of their selection.

The attention-grabbing query although is: did hive really catch any errors, or did is simply act as an additional affirmation that all the things’s all proper? And the reply is, each. Hive caught three fork-unrelated bugs in Geth, however additionally closely aided Geth’s hard-fork improvement by repeatedly offering suggestions on how modifications affected community habits.

There was some criticism of the go-ethereum staff for taking their time on the hard-fork implementation. Hopefully folks will now see what we had been as much as, whereas concurrently implementing the fork itself. All in all, I imagine hive turned out to play fairly an necessary function within the cleanness of this transition.

What’s hive’s future?

The Ethereum GitHub group options [4 test tools already](https://github.com/ethereum?utf8=%E2percent9Cpercent93&question=take a look at), with not less than one EVM benchmark device cooking in some exterior repository. They aren’t being utilised to their full extent. They’ve a ton of dependencies, generate a ton of junk and are very sophisticated to make use of.

With hive, we’re aiming to combination all the assorted scattered assessments below one common consumer validator that has minimal dependencies, may be prolonged by anybody, and might run as a part of the each day CI workflow of consumer builders.

We welcome anybody to make a contribution to the challenge, be that including new purchasers to validate, validators to check with, or simulators to seek out attention-grabbing networking points. Within the meantime, we’ll attempt to additional polish hive itself, including help for operating benchmarks in addition to mixed-client simulations.

With a bit or work, perhaps we’ll even have help for operating hive within the cloud, permitting it to run community simulations at a way more attention-grabbing scale.