feat(oft-solana): add fallback for when getSimulationComputeUnits fails

[nazreen]

Motivation

We've found that in times of network congestion, the RPC method that supports the getSimulationComputeUnits function would fail, resulting in scripts being stuck as every transaction currently relies on getSimulationComputeUnits. This blocks the deployment of Solana OFTs.

Solution

Given that as part of a Solana OFT deployment, we are executing a standard set of instructions, we can introduce a fallback by way of hardcoded CU estimates for each transaction type. The following is introduced into examples/oft-solana/tasks/solana/index.ts:

export enum TransactionType {
    CreateToken = 'CreateToken',
    CreateMultisig = 'CreateMultisig',
    InitOft = 'InitOft',
    SetAuthority = 'SetAuthority',
    InitConfig = 'InitConfig',
    SendOFT = 'SendOFT',
}

const TransactionCuEstimates: Record<TransactionType, number> = {
    // for the sample values, they are: devnet, mainnet
    [TransactionType.CreateToken]: 125_000, // actual sample: (59073, 123539), 55785 (more volatile as it has CPI to Metaplex)
    [TransactionType.CreateMultisig]: 5_000, // actual sample: 3,230
    [TransactionType.InitOft]: 70_000, // actual sample: 59207, 65198 (note: this is the only transaction that createOFTAdapter does)
    [TransactionType.SetAuthority]: 8_000, // actual sample: 6424, 6472
    [TransactionType.InitConfig]: 42_000, // actual sample: 33157, 40657
    [TransactionType.SendOFT]: 230_000, // actual sample: 217,784
}

We use the fallback in getComputeUnitPriceAndLimit:

...
export const getComputeUnitPriceAndLimit = async (
    connection: Connection,
    ixs: Instruction[],
    wallet: KeypairSigner,
    lookupTableAccount: AddressLookupTableAccount,
    transactionType: TransactionType
) => {
    const { averageFeeExcludingZeros } = await getFee(connection)
    const priorityFee = Math.round(averageFeeExcludingZeros)
    const computeUnitPrice = BigInt(priorityFee)

    let computeUnits

    try {
        computeUnits = await getSimulationComputeUnits(
            connection,
            ixs.map((ix) => toWeb3JsInstruction(ix)),
            toWeb3JsPublicKey(wallet.publicKey),
            [lookupTableAccount]
        )
    } catch (e) {
        console.error(`Error retrieving simulations compute units from RPC:`, e)
        console.log(
            `Falling back to hardcoded estimate for ${transactionType}: ${TransactionCuEstimates[transactionType]} CUs`
        )
        computeUnits = TransactionCuEstimates[transactionType]
    }
    ...

Notes:

• The transaction with the most volatile CU count is createToken as it has CPIs into the Metaplex program. More basic transactions like setAuthority have more stable CU count.
• Estimates may need to be updated if they are found to be insufficient.
• Based on errors reported recently, getSimulationComputeUnits (to estimate CU limit) is the one failing and not getRecentPrioritizationFees (rpc method to estimate CU price)

Testing

To test, init a new repo with the following:

LAYERZERO_EXAMPLES_REPOSITORY_REF=#nazreen/oft-solana/get-simulation-cu-constants LZ_ENABLE_SOLANA_OFT_EXAMPLE=1 npx create-lz-oapp@latest

To simulate the getSimulationComputeUnits method failing in times of RPC congestion, make the following temporary edit to examples/oft-solana/tasks/solana/index.ts:

    try {
       throw 'Simulating Error' // <----- add this line
        computeUnits = await getSimulationComputeUnits(
            connection,
            ixs.map((ix) => toWeb3JsInstruction(ix)),
            toWeb3JsPublicKey(wallet.publicKey),
            [lookupTableAccount]
        )

Then, run any of the Solana scripts, such as

pnpm hardhat lz:oft:solana:create --eid 40168 --program-id <PROGRAM_ID> --only-oft-store true --amount 10000000000

and observe the logs

[nazreen]

Check failing due to unrelated test:

PASS test/connection/factory.test.ts
@layerzerolabs/test-devtools-ton:test: FAIL test/node.spec.ts (10.882 s)
@layerzerolabs/test-devtools-ton:test:   TON local node
@layerzerolabs/test-devtools-ton:test:     ✕ should be available (10095 ms)
@layerzerolabs/test-devtools-ton:test: 
@layerzerolabs/test-devtools-ton:test:   ● TON local node › should be available
@layerzerolabs/test-devtools-ton:test: 
@layerzerolabs/test-devtools-ton:test:     expect(received).resolves.toEqual()
@layerzerolabs/test-devtools-ton:test: 
@layerzerolabs/test-devtools-ton:test:     Received promise rejected instead of resolved
@layerzerolabs/test-devtools-ton:test:     Rejected to value: [Error: Request failed with status code 504]
@layerzerolabs/test-devtools-ton:test: 
@layerzerolabs/test-devtools-ton:test:        7 |         const client = new TonClient({ endpoint })
@layerzerolabs/test-devtools-ton:test:        8 |
@layerzerolabs/test-devtools-ton:test:     >  9 |         await expect(client.getMasterchainInfo()).resolves.toEqual({
@layerzerolabs/test-devtools-ton:test:          |               ^
@layerzerolabs/test-devtools-ton:test:       10 |             workchain: -1,
@layerzerolabs/test-devtools-ton:test:       11 |             initSeqno: 0,
@layerzerolabs/test-devtools-ton:test:       12 |             latestSeqno: expect.any(Number),
@layerzerolabs/test-devtools-ton:test: 
@layerzerolabs/test-devtools-ton:test:       at expect (../../node_modules/.pnpm/[email protected]/node_modules/expect/build/index.js:113:15)
@layerzerolabs/test-devtools-ton:test:       at Object.expect (test/node.spec.ts:9:15)
@layerzerolabs/test-devtools-ton:test: 
@layerzerolabs/test-devtools-ton:test: Test Suites: 1 failed, 1 total
@layerzerolabs/test-devtools-ton:test: Tests:       1 failed, 1 total
@layerzerolabs/test-devtools-ton:test: Snapshots:   0 total
@layerzerolabs/test-devtools-ton:test: Time:        10.981 s

[nazreen]

checks are now passing

[ryandgoulding]

Great work! Is this ready for review?

[nazreen]

Great work! Is this ready for review?

Yes! Sorry, forgot it was created as a draft.

[ItsAdel]

left some big picture suggestions

[ItsAdel]

A few thoughts here:

I see we are overshooting the estimate for a lot of these Transaction Types, would we need to warn the users about the potential of overpaying ?

Another point here is that the estimates might not reflect the network CU usage overtime. In a few months time if everyone decides to use Solana, these values would need to be manually changed. Are we okay with making that manual change to keep the example upto date?

[ItsAdel]

thoughts on trying more than once before defaulting to estimates?

potentially using exponential backoff and retry and a couple times before defaulting

[ItsAdel]

nit: try/catch can just be condensed to

const computeUnits = await getSimulationComputeUnits(
  connection,
  ixs.map((ix) => toWeb3JsInstruction(ix)),
  toWeb3JsPublicKey(wallet.publicKey),
  [lookupTableAccount]
).catch(e => {
  console.error(`Error retrieving simulation compute units from RPC:`, e);
  console.log(
    `Falling back to hardcoded estimate for ${transactionType}: ${TransactionCuEstimates[transactionType]} CUs`
  );
  return TransactionCuEstimates[transactionType];
});