Confirmation of transaction each second

[hameleonick]

Hi all, have next problem with version 1.0.0-beta.26

When I use "methods.myMethod.send" from the contract I have subscriptions for next events: transactionHash, confirmation, error etc.
And for some reason, I received confirmation each second with a new confirmationNumber from 0 to 24. But in fact, I see that block number was not changed properly. Could you help me with this?

Please let me know if you need more details.

[frozeman]

You are using an HTTP connection?

[hameleonick]

This problem I have on the client side. I'm using additionally MetaMask Plugin for browser and I have something like re-initialisation of web3 instance:

window.web3 = new Web3Local(window.web3.currentProvider || Web3Local.givenProvider);

But it also checked with: new Web3Local(Web3Local.givenProvider)
And I had the same situation.

Thx for the response.

[kazaff]

@hameleonick me too..

In my environment(private chain), Even i had stopped miner, this confirmation event still emitted, So what this event mean?

@frozeman yep, I based on http connection.

[marwand]

I can confirm this issue occurs only with http rpc.
websocket works normally (returns only one confirmation)

Setup:
[email protected]
[email protected] (Local private rpc server, one node)

JS Code:

return MyContract.methods.MyMethod().send({from: this.accounts[1]})
.on('confirmation', (confirmationNumber, receipt) =>{
    console.log("%c Confirmation #"+confirmationNumber+" Receipt: ","color: #0000ff");
    console.log(receipt);
})
.on('error', (error) =>{
    console.log("%c Error:: "+error.message,"color: #ff0000");
    return false;
});

Output:
Receives the same receipt 25 times.

App.js:228 Confirmation #0 Receipt: {blockHash: "0x96c23b505bffebf034c648d809f0612d649e1023318bba1fe00bf7f9b2dd22cb", blockNumber: 22, contractAddress: null, cumulativeGasUsed: 225822, from: "0x069c8dc5839356fa40a4763d5deb9c6221ca22fc", …} blockHash : "0x96c23b505bffebf034c648d809f0612d649e1023318bba1fe00bf7f9b2dd22cb" blockNumber : 22 contractAddress : null cumulativeGasUsed : 225822 events : {LogNewCampaign: {…}} from : "0x069c8dc5839356fa40a4763d5deb9c6221ca22fc" gasUsed : 225822 logsBloom : "0x00000000000000000000000000000000000000000000080000000000000000000000000000000000000000000000000010040000000001000000000000000000000000000000000040000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000010000000000000000000040000000000000000000000000000000000000000000000000000000000000000000000000000002000000000000000000000000000000000000000000000000000000040000000000000000000000000000000000000000000" status : "0x1" to : "0x32eaa84f5ffb08f519ba0603f9a415066849a105" transactionHash : "0xe4b776183172e1068e0b02b5e45a234e240818fb063e1a14f77a6e4dc807b81e" transactionIndex : 0 __proto__ : Object
.
.
.
Start Campaign:: Confirmation #24 Receipt: {blockHash: "0x96c23b505bffebf034c648d809f0612d649e1023318bba1fe00bf7f9b2dd22cb", blockNumber: 22, contractAddress: null, cumulativeGasUsed: 225822, from: "0x069c8dc5839356fa40a4763d5deb9c6221ca22fc", …} blockHash : "0x96c23b505bffebf034c648d809f0612d649e1023318bba1fe00bf7f9b2dd22cb" blockNumber : 22 contractAddress : null cumulativeGasUsed : 225822 events : {LogNewCampaign: {…}} from : "0x069c8dc5839356fa40a4763d5deb9c6221ca22fc" gasUsed : 225822 logsBloom : "0x00000000000000000000000000000000000000000000080000000000000000000000000000000000000000000000000010040000000001000000000000000000000000000000000040000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000010000000000000000000040000000000000000000000000000000000000000000000000000000000000000000000000000002000000000000000000000000000000000000000000000000000000040000000000000000000000000000000000000000000" status : "0x1" to : "0x32eaa84f5ffb08f519ba0603f9a415066849a105" transactionHash : "0xe4b776183172e1068e0b02b5e45a234e240818fb063e1a14f77a6e4dc807b81e" transactionIndex : 0 __proto__ : Object

[fdouglis]

I just started hitting this problem as well. It isn't just that it reports a new confirmation on the same block repeatedly, it's that the deploy call that is hitting this doesn't return until web3 gives up by hitting its internal limit. So it takes many seconds to deploy. The weird thing is that this started happening just on one system in particular -- I had deployed on EC2, turned that into an AMI, launched a new instance from it, and hit this problem. I'm wondering if the message is a total red herring and it is actually some security policy or something else at play.

I will add I am using ganache (it happens with ganache-cli as well).

[fdouglis]

Follow up: I just came across #1393 ... seems to suggest that the new behavior may indeed be a recent, intentional change to the library. I'm wondering if my application isn't keeping up with the latest and greatest interface.

[fdouglis]

And another followup: I saw that things ran fine on a VM where I had been running beta 30, and not on a new one running 33. I removed node_packages and specified 30 specifically in my package. json, and confirmed that it had the same behavior I'd been used to. So it's definitely a recent change to web3 that broke my script.

[fdouglis]

Surprised by the radio silence on this. If the change from 30 toward 33 was unintended, I'd hope someone would identify this and agree, with the assumption the regression would be fixed at some point. If it was intentional, as #1393 suggests, can someone explain why, and suggest what the right model is to avoid sticking around?

Is the issue that in a real deployment, you'd want multiple confirmations, or something along those lines? If that is the case, then there should be a flag to limit the number needed, for use in test cases like ganache.

[cgewecke]

@fdouglis Hi. (I wrote #1393)

Tried to reproduce this on ganache-cli with 31 and it seemed to work as expected. When the PromiEvent is thenned, it immediately resolves a receipt. Then the confirmation handler fires repeatedly and also resolves the receipt. They can be used in combination. Will double-check this behavior against geth dev shortly.

If you have a chance, could you post a small example deployment call that fails against ganache-cli so I can debug?

[fdouglis]

OK, here is a pretty minimal js file that repeats the problem. Note, I said it worked in 30 and not 33. So I don't know if 31 is the one that broke it, or 32 or 33.

The Solidity file is a trivial contract that takes a single uint32 as a parameter.

[fdouglis]

BTW, there may be some misunderstanding on our part of how to handle the deploy, and maybe it's a question of documentation rather than a bug.

[cgewecke]

@fdouglis Nice script.

With these changes: (using v33)

• the default gas value set to: 0x4C4B40 (5 million)
• the config:

wsAddrGeth: 'ws://localhost:8546',
rpcAddr: 'http://localhost:8545',

Using the rpcAddr and running latest ganache-cli the output looks correct:

...etc...
Ethereum transaction 0x18a54c1ef5460d21d8238f768b580cb1bece99a9d4334885fa87392231947bc6 successfully posted.
txHash:  0x18a54c1ef5460d21d8238f768b580cb1bece99a9d4334885fa87392231947bc6
Addr:  0x636b064Cfd708114E6A906eE5F2227A96668802A
hit -then- clause of deploy
0x636b064Cfd708114E6A906eE5F2227A96668802A
confirmation callback ignored
confirmation callback ignored
...etc...

Same result using wsAddrGeth and running dockerized geth in dev mode with websockets enabled

docker run -it -p 8546:8546 -p 30303:30303 \ 
             ethereum/client-go  \ 
            --ws --wsaddr="0.0.0.0" --wsorigins "*" \
            --dev --dev.period 2 \

Output:

... etc ...
Ethereum transaction 0x98061db4e759ce32ebbbfdcb68037379a149be7cee6b25a37e1abe5619eedc92 successfully posted.
txHash:  0x98061db4e759ce32ebbbfdcb68037379a149be7cee6b25a37e1abe5619eedc92
confirmation callback ignored # <-- note (different than ganache) 
Addr:  0xC52C92F5e9C40d81B5Ca2547DF1878Eb8D4d53E8
hit -then- clause of deploy
0xC52C92F5e9C40d81B5Ca2547DF1878Eb8D4d53E8
confirmation callback ignored
confirmation callback ignored

Same result using rpcAddr and running dockerized geth in dev mode with rpc enabled:

docker run -it -p 8545:8545 -p 30303:30303 \
            ethereum/client-go \ 
            --rpc --rpcaddr "0.0.0.0" \
            --dev --dev.period 2 \

Does this work on your end?

[fdouglis]

@cgewecke, thanks. When you say "the output looks correct" what do you mean? With v0.30, I would see a single confirmation, and then it would return. Now I see many, and it doesn't return until they stop. This slows down interactions that are deploying the contract and waiting for it to be deployed.

It may be that this is an artifact of how we developed the app. That if you deploy in something that continues to run other threads and do useful work, you can get these confirmations and drop them on the floor and not notice.

[cgewecke]

@fdouglis The promise is resolving and the emitter keeps emitting. Isn't this correct behavior?

hit -then- clause of deploy

[Edit] What is your execution context? The above seems like it would be unproblematic in node or the browser.

[fdouglis]

Well, this is what I've been trying to figure out. It appears you explicitly wanted to change things to have this behavior, but it was unexpected from our perspective. Why would the same deploy keep getting the same confirmation many times? I could see that if, for instance, I deployed it into a chain with many nodes and each one somehow confirmed it. Here, it's just ganache. So, in the past, I used to see that it was deployed -- I'm actually not sure I got the confirmation debug message at all, but certainly not many times.

And, to be clear, the fact that I see the callback repeatedly wouldn't be a huge issue under some circumstances. I don't have to log each such occurrence (and in fact at some point I tweaked my code to print the 2nd then suppress any further). But it's the delay that is the problem. We run this in a bit of a synchronous form: invoke deploy, wait for it to finish, and save the contract address in a file. Subsequent references use the address from the file if we start nodejs again. So if there is a lag because it's not really really done until it hits the limit on these confirmations, there is a problem.

If the answer is, that's the expected behavior, even though we happened to use a version that didn't have that behavior and got away with it, we need to decide how to deal. For instance, detecting the successful deployment and returning immediately even if more confirmations may arrive in the background.

[cgewecke]

@fdouglis Ah I understand. There's some discussion of the necessity of repeated confirmations from a chain architecture perspective here. Additionally there are use-cases relating to Infura or any other load balanced public node clusters, nicely articulated here

[fdouglis]

OK, @cgewecke, so this gets back to my comment about testing versus the real world. I could find no documentation suggesting that number of confirmations is configurable. I fully agree you would want to count your confirmations in real deployments. But when you are talking to ganache, you're talking to a single blockchain, so you're simply repeating the same confirmation, right? A parameter in truffle.js would handle this case. Perhaps it's already there, just not easily found, or perhaps it could be added?

[fdouglis]

P.S. I could see it if there were some benefit to not just knowing how many nodes mined a contract but also how many blocks on one node were written afterwards, with the idea being that the deeper you get, the more reliable the transaction. However, in my test, I was deploying the contract, then waiting for the receipt, then moving on. Nothing else was happening, hence the appearance that the same confirmation was being repeated numerous times (which sounds like a bug from where I sit). And since it didn't return from the full deploy sequence until enough confirmations were hit, it just needlessly slowed down deployment.

To summarize,

• Configurable number of confirmations before proceeding
• Is it a bug to get the same confirmation multiple times from the same node when no new blocks have been mined?

[cgewecke]

@fdouglis Fair points!

Is it a bug to get the same confirmation multiple times from the same node when no new blocks have been mined?

Yes, I believe that is the bug noted in this issue, caused by polling at 1 second intervals. It affects non-websocket client connections.

[fdouglis]

I see. I think I was confused because when you showed your output with all the confirmations, you seemed to be suggesting it was working as expected. Thanks.

[dpurhar27]

Sorry @fdouglis have you found a way to fix this? My DApp was functioning correctly until 2 days ago, and now my front-ends event.watch() appears to not see the event emitted until I run the function again (by clicking the button on the gui again). As soon as I push the button, the event.watch() sees the event from the previous turn. Is this related to the problems you are/were experiencing? I also run this in a synchronous fashion, I click the button, wait for the event, then continue onto compare what the event returns. So it is completely messing the DApp functionality now.

[fdouglis]

@dpurhar27 Not sure if its' the same issue or not, sorry. I solved it by either fixing on an earlier beta or more actively ignoring the extra confirmations (I'd been printing a message, and stopped). The latter is quieter but still has performance issues.

I've asked that they add a parameter to override this, but also, they seemed to say it's a known bug with the HTTP interface, if that's what you're using.

[CryptoKiddies]

I can confirm this issue connecting via http. As soon as the contract is mined, I get a confirmation fired every second up to 24. Is there a proposed fix?

[HugoPeters1024]

@GeeeCoin I encouter exactly the same

[frozeman]

Sorry for the late reply. It certainly would be the polling if the HTTP, as event checking is here different. I’m currently not able to look into this. But a PR is welcome. It would be in the web3-core-method package where things need a fix

[nivida]

Yes the "newHeads subscription" gets fired each second with the HTTP provider because it does not support subscriptions. If you're using and WebSocket or IPC provider it uses the real "newHeads" subscription and then it will be triggered on each head. But yes it's wrong that it increases the confirmation counter.

This should be fixed with the PR #2000