-
Notifications
You must be signed in to change notification settings - Fork 89
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
feat: added packet count telemetry metrics #2797
Changes from 2 commits
6e29243
720a702
38c82d1
24c5dbf
023d643
3bc58d7
b958b34
17852af
f5b5bf4
cffe639
1c29f9d
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,3 +1,43 @@ | ||
# `core` | ||
|
||
> Toolkit comprised of core services and middleware needed to develop ILP applications. | ||
|
||
## Telemetry Collection | ||
|
||
Currently, we collect packet count and packet amount metrics directly within the connector core. These metrics are captured at the Interledger layer to track packet activity while ensuring privacy by collecting amounts at the packet level, rather than at the transaction level. This approach helps to preserve privacy, as we do not expose entire transaction amounts, while also incorporating privacy-preserving measures into the collected amounts. You can read more about privacy [here](https://rafiki.dev/telemetry/privacy/). | ||
|
||
### Why We Collect on the Sending Side | ||
|
||
The first decision in collecting this data was whether to do so on the sender's side or the receiver's side. We opted for the sender’s side to maintain consistency across our metrics, particularly for calculating average transaction amounts and times, which are tied to the outgoing payment flow in the Open Payments (OP) layer. By collecting metrics from the same perspective, we ensure alignment with how other metrics (like transaction completion times) are captured. | ||
|
||
We also considered collecting metrics on both the sending and receiving sides, capturing metrics when prepare packets were received by the receiver and when fulfill or reject responses were received by the sender. However, this could lead to unreliable metrics, as telemetry is optional and might not be enabled on all nodes. | ||
|
||
### Why We Chose handleIlpData | ||
|
||
Given these considerations, we decided to place our packet count and amount metrics in the `handleIlpData` function within the Rafiki Connector Core. This function plays a crucial role in processing ILP packets, handling both outgoing payments and quotes. | ||
|
||
The specific metrics we collect here include: | ||
|
||
- packet_count_prepare: Counts the prepare packets sent, collected before the middleware routes are executed. | ||
- packet_count_fulfill: Counts the fulfill packets received, collected after receiving a reply from the receiver. | ||
- packet_count_reject: Counts the reject packets received. | ||
- packet_amount_fulfill: Records the amount sent in fulfill packets. | ||
|
||
These metrics provide valuable insights, including potential packet loss. | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. I was thinking about this, and this is not exactly packet loss like it is defined in the networking sense (since we only count them in the originating connector) , but something like rate of success/error instead |
||
|
||
### Challenges with the Current Setup | ||
|
||
While `handleIlpData` is an effective location for telemetry collection, it has some limitations: | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. I would prefer not having function names in documentation since it becomes outdated quite quickly, I think we can just remove references to it and keep just the general explanation which you have There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. The reason I've included function names is try and avoid any duplicated work when the next person needs to update this or change the metric location. Is there somewhere better for me to use function names and get really specific outside of the README? |
||
|
||
Sender-Side Limitation: Currently, metrics are only collected on the sender side. This is adequate for now but does not capture data from connecting nodes, which we plan to address in future implementations when multi-hop support is added. | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Might be good to add a short explanation for the connector nodes: eg in A > B > C, we only collect packet information in A |
||
|
||
### Other Considered Locations for Metrics Collection | ||
|
||
We explored several alternative locations for collecting telemetry metrics within the Rafiki Connector Core: | ||
|
||
- Dedicated Middleware: Initially, we implemented a dedicated telemetry middleware. However, this approach resulted in data being duplicated on both the sender and receiver sides, leading to inaccurate metrics. To address this, we would need to filter the data to ensure metrics are only collected on the sender side. Additionally, the telemetry middleware would need to effectively handle errors thrown in the middleware chain. To achieve this, it might be necessary to place the telemetry middleware right before the error-handling middleware, allowing it to catch and reflect any errors that occur. This would involve collecting the prepare packet count, wrapping the `next()` function in a try-catch block to capture any new errors, and then collecting reply and amount metrics after `next()` has resolved. | ||
- `ilpHandler` on the Receiving Side: We also considered adding telemetry to the `ilpHandler` function, which processes middleware on receiving and connecting nodes. However, this would lead to a fragmented metric collection, with some metrics gathered on the receiver side and others on the sender side. This fragmentation could complicate the handling of concepts like transaction count, which would require dual collection on both sender and receiver sides. We'd also have to watch for the possibility of data duplication on the connectors because their middleware might trigger in each direction, as they receive prepares and again as they receive responses. | ||
|
||
### Moving Forward | ||
|
||
As we implement multi-hop capabilities and further refine the connector architecture, we will likely identify better entry and exit points for telemetry collection. This will allow us to capture a more comprehensive set of metrics across both sender and receiver nodes, ensuring a complete and accurate understanding of the network’s behavior. |
This file was deleted.
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,61 @@ | ||
import { TelemetryService } from '../../../../telemetry/service' | ||
import { IlpResponse } from './middleware/ilp-packet' | ||
import { ValueType } from '@opentelemetry/api' | ||
|
||
export function incrementPreparePacketCount( | ||
unfulfillable: boolean, | ||
prepareAmount: string, | ||
JoblersTune marked this conversation as resolved.
Show resolved
Hide resolved
|
||
telemetry: TelemetryService | ||
): void { | ||
if (!unfulfillable && Number(prepareAmount)) { | ||
telemetry.incrementCounter('packet_count_prepare', 1, { | ||
description: 'Count of prepare packets that are sent' | ||
}) | ||
} | ||
} | ||
|
||
export function incrementFulfillOrRejectPacketCount( | ||
unfulfillable: boolean, | ||
prepareAmount: string, | ||
JoblersTune marked this conversation as resolved.
Show resolved
Hide resolved
|
||
response: IlpResponse, | ||
telemetry: TelemetryService | ||
): void { | ||
if (!unfulfillable && Number(prepareAmount)) { | ||
if (response.fulfill) { | ||
telemetry.incrementCounter('packet_count_fulfill', 1, { | ||
description: 'Count of fulfill packets' | ||
}) | ||
} else if (response.reject) { | ||
telemetry.incrementCounter('packet_count_reject', 1, { | ||
description: 'Count of reject packets' | ||
}) | ||
} | ||
} | ||
} | ||
|
||
export async function incrementAmount( | ||
unfulfillable: boolean, | ||
prepareAmount: string, | ||
response: IlpResponse, | ||
code: string, | ||
scale: number, | ||
telemetry: TelemetryService | ||
): Promise<void> { | ||
if (!unfulfillable && Number(prepareAmount)) { | ||
JoblersTune marked this conversation as resolved.
Show resolved
Hide resolved
|
||
if (response.fulfill) { | ||
const value = BigInt(prepareAmount) | ||
await telemetry.incrementCounterWithTransactionAmount( | ||
'packet_amount_fulfill', | ||
{ | ||
value, | ||
assetCode: code, | ||
assetScale: scale | ||
}, | ||
{ | ||
description: 'Amount sent through the network', | ||
valueType: ValueType.DOUBLE | ||
} | ||
) | ||
} | ||
} | ||
} |
This file was deleted.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
thanks for adding this