START
⇓
releaseMempool ┌───────────────┐
╭─────▶│ Idle │⇒ DONE
│ └───────┬───────┘
│ │
│ │ acquireMempool
│ │
│ │
│ │ (re)acquireMempool
│ │ ╭───────╮
│ ▼ │ │
│ ┌───────────┴───┐ │
╰──────┤ Acquired │◀──╯
└───┬───────────┘
│ ▲
nextTransaction │ │
hasTransaction │ │
sizeOfMempool │ │
╰───────╯
To inspect the node’s local mempool, one may rely on the mempool monitoring mini-protocol. This protocol provides way to list all transactions sitting in the mempool, but also, to query the size of the mempool, the number of transactions currently in the mempool as well as the current capacity (based on network parameters).
As for the other mini-protocols, the mempool monitoring is a stateful protocol with explicit state acquisition driven by the client. That is, clients must first acquire a mempool snapshot for running queries over it. Once acquired, queries are guaranteed to be consistent. In particular, nextTransaction will never yield twice the same transaction for the same snapshot and sizeOfMempool will remain constant.
acquireMempool is a blocking call. The server will only reply once a “new” snapshot is available. “New” means different from the currently acquired snapshot. Seemingly, the first acquireMempool is instantaneous. This allows for clients to passively wait for changes without active polling. A typical pattern of usage would be to acquire a snapshot, list all transactions from the mempool via nextTransaction and then, block on acquireMempool for a change; then repeat.
First, client must always acquire a snapshot and hold onto it for subsequent queries. To list all queries, one must call nextTransaction repeatedly until it yields null. So for instance, if the mempool currently contains three transactions t0, t1 and t2, one can list all transactions from the mempool via the following sequence (schematically):
1. acquireMempool → acquireMempoolResponse 2. nextTransaction → nextTransactionResponse t0 3. nextTransaction → nextTransactionResponse t1 4. nextTransaction → nextTransactionResponse t2 5. nextTransaction → nextTransactionResponse null
Another option is simply to query for a specific transaction via hasTransaction, which yields True or False depending on whether the transaction is currently in the mempool or not.
1. acquireMempool → acquireMempoolResponse 2. hasTransaction t0 → hasTransactionResponse True 3. hasTransaction t1 → hasTransactionResponse True 4. hasTransaction t5 → hasTransactionResponse False
At any moment, it is also possible to interleave a sizeOfMempool query to get the acquired snapshot’s size (in bytes), number of transactions and capacity (in bytes).
The capacity refers to the maximum size of the mempool. It is currently defined as twice the network block size and can be adjusted via protocol updates.
Since 5.3.0, Ogmios can also return full transactions as a result of nextTransaction. This must be however explicitly requested from clients by providing an extra (optional) argument to each nextTransaction request:
{
"jsonrpc": "2.0",
"method": "nextTransaction",
"params": {
"fields": "all"
}
}
"fields" accept only one value ("all") and can be omitted. When present, the result from the response will contain a full transaction. When omitted, result will only contains a transaction id.
Some important notes to keep in mind regarding the management of the mempool:
This protocol gives access to transactions that are submitted locally, by the connected client via the transaction submission protocol. In case of block producing nodes (i.e. stake pools), transactions pulled from peers may also be available.
The protocol does not guarantee observability of all transactions passing through the mempool. There’s an inherent race condition between the client acquiring snapshots and the node managing it internally. Thus, while a client is holding a snapshot, it may still submit transactions through the transaction submission protocol, which may be accepted, processed and included in the ledger before the client next’s acquireMempool. So, it is possible for clients to miss transactions passing through the mempool should they be concurrently submitting them.
Furthermore, while the presence of a transaction in the mempool qualifies it as pending, the absence of transactions in the mempool does not guarantee their inclusion in the ledger (transaction may be discarded from the mempool for various reasons). In particular, a valid transaction may leave the mempool to be included in a block which later result in a lost fork (e.g. because of a lost slot battle) and may never end up in the ledger. The node does not automatically re-insert transactions into the mempool.
const WebSocket = require('ws');
const client = new WebSocket("ws://localhost:1337");
// Helper function
function rpc(method, params) {
client.send(JSON.stringify({
jsonrpc: "2.0",
method,
params
}));
}
client.on('message', e => {
const next = JSON.parse(e).result;
if (next.transaction === null) {
client.close();
} else {
console.log(next.transaction);
rpc("nextTransaction");
}
});
client.once('open', () => {
rpc("acquireMempool");
});
Errors from the mempool monitoring protocol are in the range 4000-4999 and are listed below.
The complete description of the mempool monitoring requests and responses can be found in the API reference.
Plus, test vectors are available on the repository for testing, debugging and to serve as examples.