darwinia-network / bridger Goto Github PK

bridger should receive an answer from shadow about its status such as its latest block when error. not just pop up a json error.

[2021-01-05T09:05:55Z ERROR darwinia_bridger::service::relay] Error {
        context: "Fail to get parcel from shadow when affirming ethereum block 9401625",
        source: Error {
            context: "Fail to parse json to MMRRootJson: https://testnet.shadow.darwinia.network/ethereum/parent_mmr_root/9401626",
            source: reqwest::Error {
                kind: Decode,
                source: Error("missing field `mmr_root`", line: 1, column: 49),
            },
        },
    }

If set darwinia start back, it will re sign and send root or new authorities extrinsic.

Run bridger in daemon mode

> bridger run -d

Show running status

> bridger status
ID      SERVICE NAME                            STATUS
1       ethereum to darwinia relay              runing
2       ethereum to darwinia redeem             stopped
3       ethereum to darwinia guard              stopped
4       ethereum tracking                       runing
5       darwinia tracking                       runing

Start & stop service

> bridger start 2

> bridger stop 2

Logs

 𝝺 ./target/debug/bridger keys 
sudo key: d43593c715fdd31c61141abd04a99fd6822c8558854ccde39a5684e7a56da27d (5GrwvaEF...)
technical committee members: [d43593c715fdd31c61141abd04a99fd6822c8558854ccde39a5684e7a56da27d (5GrwvaEF...)]
 𝝺 ./target/debug/bridger guard 
[2020-10-26T16:00:16Z INFO  darwinia_bridger::cmd::guard] Init API succeed!
[2020-10-26T16:00:16Z ERROR darwinia_bridger::cmd::guard] Bridger("Guard service is not running because the relayer is not a member of the technical committee!!!")

And why using !!! here? A bridger is not a technical committee that might be a common situation.

Could cause redeem service stop running?

• Relay Pending Header Event
• Relay Guard Checking Event. (Vote Aye/Nay)
• Relay Confirmed Header Event.
• Backing Redeem Event (print origin)
• Relayer Game Events

For example, change the rpc in [eth] to a non-existent node address:

# ...
[eth]
rpc = 'https://mainnet.infura.ioabcd/v3/...'
start = 11149205
# ...

Bridger should report an error if it's not able to connect to the rpc endpoint. Please do not check the address ourselves, instead, it is better just throwing out the error from the HTTP client, as this is just an example case to let you easily reproduce the issue. We have tons of possibilities of different errors, like connection refused if the port is misconfigured, no route to the host if the IP is misconfigured, or missing CA certificates if we don't have ca-certificates installed, etc...

Desc

The error!() should print logs actually out of our expectations, and the fail of relayable check are in expect, they should be warn!(), different from the real error.

A refactor is required to do later.

In current chain protocol, headers must be confirmed increasingly.

This approve extrinsic will fail, these pending headers will be cleaned by system in confirm end block.

Prevent transactions with the same nonce from being sent

There were redeem examples, recent example is guard:

b8ff863

When handling the Darwinia event, this error will appear repeatedly, but it will not cause a problem.

[2020-12-30T13:19:19Z ERROR darwinia_bridger::service::subscribe] Encounter error when handle events of block 378287: Scale codec error: No such variant in enum Phase

Lines

Desc

The darwinia-bridge-primitives crate already implemented the same function

Currently, there is a relayer key in config which is only used when proxy account is open.

A minor suggestion would be.

• rename relayer to a more accurate name, e.g. real
• adding a proxy tag to define keys only for proxy
• current seed key have different meanings in two cases, if proxy is closed, it means the private key of relayer, if proxy is opened, it means the private key of proxy account.
• Adding documents explain the config. (@clearloop said that it is auto generated now, but it is not friendly to add docs for it, prefer removing auto generate mechanism)

seed = ""

[proxy]
real = ""

• Move start to outside of config, this should not hardcoded in config and config is not for data store.
• Provide a command for use to query/change the current ethereum scan start block.

e.g.

They have different proxytypes:

Crab: https://github.com/darwinia-network/darwinia/blob/master/runtime/crab/src/lib.rs#L738

Darwinia: https://github.com/darwinia-network/darwinia/blob/master/runtime/darwinia/src/lib.rs#L709

• Configuration
• Shadow Service
• Operation
• Logging
• Proxy Document, #15

In the startup information, add whether the current relay account is a sudo account or a member of the technical committee.

Since I'm using Kubernetes to deploy the bridger, probably adding the version to logs at the time when bridger launches would be nice. That way, operators like me will be able to know the version of current running bridger, by just inspecting log output without running another command in the container.

Desc

Could not build futures-macro using sccache wrapper.

Solution

Stop the sccache backend server solved the problem on my machine temporarily

Part of the logs:

darwinia-bridger-ws-codec-reserved-op-code.log

The size of full logs is about 6 GiB, so I'd have to upload only a part of it. I believe those left logs contain no valuable info but only the error "Client Error: Inner(Ws(Codec(ReservedOpCode)))".

• change to technical committee
• test

If mmr is 0x0, it means shadow is not ready yet, maybe guard should wait for a while to vote.

In another aspect, shadow service's real time performance is important for guard to make decision.

if some error occurs here, we may lose events in this block since the next block index still moves forward.

• Check #154

It would be great if we can print out the value of "invalid URI", that helps while debugging or I'd have to look into the config file to find out where's the invalid character.

Desc

This will make the checking of pull requests easier, for example, rustfmt will :

• Order the dependencies
• Unify the code style of different developers

Lots of editors will run rustfmt on saving changes.

Husky

This still can be implemented by using husky.

Desc

Our contract addresses are rarely changed, it might be a choice to write it inside and switch by flag, for example bridger run -v --testnet

Or a switcher like the proxy config:

testnet = false;

• affirm, when received request of header relay into light client, and there are no existing games match requirement (game block number is smaller than the requested block number).

• dispute_and_affirm, when received affirm event and have different checking result.

• extend_affirmation, when received event of new round, and continue to follow the right affirmation if necessary. (make sure that last_leaf is using first_affirm_header_number - 1, because protocol is using mm_root in first affirmed header parcel to verify consistency)

• Check #44

First release and tag,

then ci auto publish on crates.io

• proxy support: #16
• modify proxy config
• add proxy pallet to node-template
• test with node-template

Now the tx data center is a memory pool, if the program stopped, the tx data will be lost.
When the user restarts the program, he must reset the start param in the config file, or the program will do repeat work.