Skip to content
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

Improve APM spans (no more <anonymous>) #1267

Merged
merged 5 commits into from
Apr 4, 2024
Merged

Conversation

timotheeg
Copy link
Contributor

@timotheeg timotheeg commented Apr 3, 2024

Problem

We have a lot of spans in APM traces which are <anonymous>. While not a deal breaker to get value out of APM, it is annoying.

image

The anonymous functions are due to the way some methods are declared in classes, which is done to be able to specify types.

e.g. in ReviewsRouter.ts, the assignment syntax creates anonymous methods

  compareDiff: RequestHandler<
    { siteName: string },
    { items: EditedItemDto[] } | ResponseErrorBody,
    unknown,
    unknown,
    { userWithSiteSessionData: UserWithSiteSessionData }
  > = async (req, res) => {
    // ... 
  }

while in collectionPages.js, this style of named method works:

  async updateCollectionPage(req, res, next) {
    // ...
  }

Note: The first style of method declaration (assignment) does NOT create methods on the class' prototype. The methods are created on the instance themselves at construction (and so just FYI the call to autoBind(this) are actually useless for classes which only use that style of declaration)

TODO: We should ensure that all methods we use as express route handlers are named to have nice usable traces

Solution

  • Provide a utility nameAnonymousMethods() to ensure all methods of an instance and its class are named (invoked at construction)
  • Update all router classes (ts files) to ensure all methods are named

Updating the Router classes was done via the script below, which was then cleaned up on commit by eslint/prettier:

grep 'autoBind(this)' $(git grep Router | grep class | grep Router | cut -d: -f1) | cut -d: -f1 | grep -v -E '.js$' | while read f; do
  echo 'import { nameAnonymousMethods } from "@root/utils/apm-utils"' > temp;
  sed -E -e 's/autoBind\(this\)/nameAnonymousMethods(this);autoBind(this)/' "${f}" >> temp;
  mv temp "${f}";
done

Verified on staging:
image

Tests

  • Load the CMS and perform some actions, all happy paths should work
  • Find a span for a Review compare, and verify the span name is compareDiff (NOT <anonymous>)

@timotheeg timotheeg requested a review from a team April 3, 2024 06:44
@timotheeg timotheeg force-pushed the improve_apm_spans branch from fd24d85 to 65e92e2 Compare April 3, 2024 07:03
@@ -150,6 +151,7 @@ export class FormsgSiteLaunchRouter {
this.usersService = usersService
this.infraService = infraService
// We need to bind all methods because we don't invoke them from the class directly
nameAnonymousMethods(this)
autoBind(this)
Copy link
Contributor Author

@timotheeg timotheeg Apr 3, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Note on this for reviewer: autoBind(this)in this file (and all the router files thathave been converted to typescript) doesn't actually do what we think it does 😑😢

When the methods are declared like this in the class:

 handleSiteLaunchFormRequest: RequestHandler<
    never,
    string,
    { data: { submissionId: string } },
    never,
    { submission: DecryptedContentAndAttachments }
  > = async (req, res) => {
   //. ...
}

then they are:

  1. anonymous
  2. NOT added to the class' prototype (so autoBind() will not get to them at all)
  3. Added to the instance directly and bound

@alexanderleegs alexanderleegs merged commit ccde289 into develop Apr 4, 2024
12 checks passed
@mergify mergify bot deleted the improve_apm_spans branch April 4, 2024 05:05
This was referenced Apr 4, 2024
alexanderleegs added a commit that referenced this pull request Apr 4, 2024
* fix(server): server should die if unable to connect to db (#1265)

## Problem

something in me wanted to check if we indeed exit if we fail to connect to db, and the answer is... no 


[node.js](https://nodejs.org/docs/latest-v18.x/api/process.html#processexitcode_1) states that 
> A number which will be the process exit code, when the process either exits gracefully, or is exited via process.exit() without specifying a code.

so it does not actually do the exiting, which leads to silent failures

* fix(dig): dig not working (#1246)

## Problem

We do have some checks in the backend to check prior to any site launch that there are AAAA records/CAA records present. In this case, the dig commands to check the AAAA records [failed](https://ogp.datadoghq.com/logs?query=service%3Aisomer%20env%3Aproduction%20%22An%20error%20occurred%20while%20performing%20dig%20for%20domain%3A%22%20&cols=host%2Cservice&fromUser=true&index=%2A&messageDisplay=inline&refresh_mode=sliding&storage=hot&stream_sort=desc&viz=stream&from_ts=1711457521321&to_ts=1711543921321&live=true).

The reason for above is that we were using a library called `node-dig-dns`. This called the dig command [directly](https://github.com/StephanGeorg/node-dig-dns/blob/master/src/index.js#L78) at a system level. However, our docker container does not have the dig command out of the box. This resulted in the existence of AAAA records not being caught.


We are also codifying a check for CAA records and ensuring that if there exist at least one  caa record and it uses our redirection service, it should have letsencrypt as one of the caa record. 

To prevent accidental commits to live indirection repo during dev, also adding a check to only commit to the indirection repository iff it is in prod environment. 
 
## Solution

just use node's dns resolver directly. this way we dont have to depend on an external library's implementation of node and dont have to install unnecessary deps in the docker.

remove dep introduced in #1244 

**Breaking Changes**

<!-- Does this PR contain any backward incompatible changes? If so, what are they and should there be special considerations for release? -->

- [ ] Yes - this PR contains breaking changes
  - Details ...
- [x] No - this PR is backwards compatible with ALL of the following feature flags in this [doc](https://www.notion.so/opengov/Existing-feature-flags-518ad2cdc325420893a105e88c432be5)

**Features**:
<img width="1493" alt="Screenshot 2024-03-27 at 11 21 44 PM" src="https://github.com/isomerpages/isomercms-backend/assets/42832651/73edd437-492b-4bff-86c9-3392cb40fe49">


### Manual test (not to be copied over to deployment notes)

- [ ] add these lines of code at the end of server.js 
```
const formResponses = [
  {
    submissionId: "",
    requesterEmail: "[email protected]",
    repoName: "kishore-test-dev-emil",
    primaryDomain: "google.com",
    redirectionDomain: "www.google.com",
    agencyEmail: "[email protected]",
  },
]

formsgSiteLaunchRouter.handleSiteLaunchResults(formResponses, "test")
```  

- [ ] Assert that the email comes out to

* fix: remove unecessary join and site retrieval (#1268)

* Improve APM spans (no more <anonymous>) (#1267)

* refactor: rename wrapper based on original function, only when original.name exists

* feat: utility to name all methods of an object

* feat: ensure all route handlers are named

* feat: drop the bound prefix in span names

* chore: remove unused eslint rule disabling

---------

Co-authored-by: Alexander Lee <[email protected]>

* fix: external links in top level nav (#1272)

* chore: bump version to v0.76.0

---------

Co-authored-by: Timothee Groleau <[email protected]>
Co-authored-by: Kishore <[email protected]>
@timotheeg timotheeg mentioned this pull request Apr 8, 2024
This was referenced Jun 27, 2024
@dcshzj dcshzj mentioned this pull request Jun 27, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants