Fail with gRPC Proxy executing Trx FrontEnd and don't redirect to another one

[AdrianKBL]

Description

After 2555 NFTs minted in Mainnet (255 Batches and each batch is based on: Transaction sending Hbar + Transactions of Minting NFTs, in total 510 Transactions has been executed), directly from the browser (It means the Transaction don't get executed in BackEnd), It gets executed directly in the FrontEnd via gRPC + Proxies, we got a 503 in the Node04 of Swirlds (We have a work around to retry 3 times to get the receipt, although I thought you guys fixed this last time we opened the issue. As far as I thought the SDK would directly rebalance to an operative Proxie & Node, but as I can see It's not working as expected. After my 3 retries the execution is cut by throwing an Error.

Also after this, we got the same Error than we reported 2-3 months ago, that you guys weren't able to reproduce it.
Transaction must have been frozen before calculating the hash will be stable, try callen freeze

This is something critical for Kabila App, since we're providing this services and our customers are having troubles with it.

Steps to reproduce

let mintTx = new TokenMintTransaction()
    .setMaxTransactionFee(new Hbar(maxFee))
    .setTokenId(tokenId);
    .setMetadata(metadata);
   mintTx.freezeWith(client);
   const mintTxSign = await mintTx.sign(supplyKey);
   const mintTxResponse =  await mintTxSign.execute(client);
   const receipt = await mintTxResponse.getReceipt(client);
    
   let txTransfer = new TransferTransaction()
     .setMaxTransactionFee(new Hbar(maxFee))
     .setTransactionMemo(memo)
     .addHbarTransfer(account1, hbarAmount.negated()) 
     .addHbarTransfer(account2, hbarAmount);
   const mintTxResponse = await txTransfer.execute(client);
   const receipt = await txSubmit.getReceipt(client);

Additional context

No response

Hedera network

mainnet

Version

2.20.0

Operating system

Windows

[ochikov]

Hello @AdriMsM91,
I've spent the weekend investigating the two bugs reported above:

• About the bug reported for retries when we hit the Node with partial outage (Node 4 with NodeAccoundID 7), I am applying a couple of screenshots with the retrying visualised (transfer and receipt retries):

You can see from the images provided that no matter the error, by default we've got 10 max retries and we are retrying and getting the receipt. On some of the screenshots we are retrying 2-3 times, and on some 5-6 times. The screenshots are taken from React App and Node.js App.

• About the second problem Transaction must have been frozen before calculating the hash will be stable, try callen freeze I was able only to reproduce it randomly during the weekend. It starts to happen only when my laptop is going to sleep or I left it unattended and the display is off. I tried a large number of times and this is the way to reproduce it.

I will continue to dig it today.

[ochikov]

@AdriMsM91 regarding the retries, I cannot see any problem. Did you lock the nodes on your side?

[AdrianKBL]

What do you mean by lock the nodes?

[ochikov]

@AdriMsM91 I mean by setting the setter:

          .setNodeAccountIds([
            new AccountId(7),
            new AccountId(8),
            new AccountId(9),
          ])

for example with specified nodes?

[AdrianKBL]

Nop,

let client = Client.forNetwork(network);

[ochikov]

Perfect, so from my applied screenshots, you can see that it works as expected. I am still trying to identify why

Transaction must have been frozen before calculating the hash will be stable, try callen freeze

is happening. Because when it occurs, all transactions are failing. I still cannot reproduce it without closing the lid of the laptop. It's super strange behavior.

[AdrianKBL]

It usually works as expected, but some times It doesn't and I don't know why.

[ochikov]

@AdriMsM91 is there any chance when you are waiting for the NFTs to be minted and for the transfer transaction, something to kill the process of the SDK? What time is needed for the execution of the process?

[AdrianKBL]

I don't think so @ochikov sometimes we have minted 8,000 NFTs without any problem in Mainnet, sometimes we have problems with 2.5k and the process is exactly the same, also differents customers have the same problem.

[ochikov]

@AdriMsM91 I cannot catch any problem related with the healthy/unhealthy nodes.

[AdrianKBL]

Yeah it's super hard to catch up, only happens when you mint really big amounts + big amounts of Hbar Transfer in Mainnet in the FrontEnd directly, without using Node.js for anything related about executing transactions. :(

[AdrianKBL]

@ochikov Again today another customer with the same problem.

[ochikov]

@AdriMsM91 I understand, but the behavior is really hard to be reproduced. I cannot propose any solution right now.

[ochikov]

@AdriMsM91 Hello, I will close the issue.

[ochikov]

Version 2.26.0-beta.2 should fix the issue.