Best Practices for CAP Node.js Apps

 

• 1. Dependency management

Issue 1.1: Suddenly my apps stopped working in production because of some updates!

Reason

Solution: Manage your dependencies use package-lock.json to freezes versions

Checking outdated by the comment: npm outdated

To update package outdated using: npm update

Recommend to use the specific version on the package

Issue 1.2: Sometimes, my app behaves in a very unexpected way

Don't add an unnecessary try-catch. 

If you don't know the Customer always an array, then you always know that there will be a for each method. So, don't need to add array check like that

 You should know data structures in your codebase. 

If you know that some customers don't have a payments property then you should check it using IF condition to check don't you try-catch around it.

DON'T

DO

Never catch Programming errors 

- They have to be fixed

- For unknown programming errors, the app must crash, fail loudly, then fix

• 2. Error handling

- Always handle operational errors: Let's just say you have here an express handler and you execute some HTTP request that means an external system is involved and whenever you have external systems or user input or whatever outside your program then they will always go wrong at some point in time. There's not a single external system that is available all the time maybe the connection is broken, or the external function error...etc. So, always make sure to handle those errors

In this case, this function could be throw in if the system is not available so you have to catch it, no other way to do it if excecueHttpRequest throw some time you can get the status of the response then you can check on httpResponse  but if this functions throws then you have to catch it then log it and send status 502 for the bad gateway for example.

CAP's remote service API does this automatically if you use CAP and you have here a handler to read books and you take external services (extSrv)  and you run this query on it and we find out that this external service is not there then we automatically that gateway so you don't have to deal with its income.

Issue: After some point in time, my app stops working properly.

If you look at the code this is the express middleware to catch all errors that means you catch and you can see it because it has four Argument(err,req, res, next) this way you catch all errors which happen during runtime and just send status 500 to the caller which sounds reasonable at first glance. But the problem is that all arrows are caught, which means you are also catching unexpected errors in programming.

Reason: Unexpected errors are always caught

Handle CAP server crashes when there are programming errors

CAP server also crashes when there are programming errors so we have to central error handler but we always check that if it's one of those 

       err instance of  TypeError ||

err instance of  ReferenceError||

err instance of SyntaxError ||

err instance of  RangeError ||

err instance of  URIError||

 Issue: There's an error in my app, but I can't find the root cause

This is very common something like this, you get a stack trace with some error that happened, and you don't know where has error happened

You try to find the error on line 3 column 9 then you look at the code and it's something like this

The reason: The error lose information when you try to re-throwing errors

Solution:Using object assign some new property like that

Then you will see an error like this with the original error at the importantFunction Then we can figure out what is the root cause.

There's also new feature in V8, V9

• 3. Transaction handling

Issue: After the first request, my app doesn't respond anymore!

This is a very common issue. Take a look at the code below it to seem good but something mission as an argument of tx function you don't provide the request object to tx

Reason: cds.tx() start a new (unmanaged) transaction which you have to commit and roll back manually.

Solution:  You have to provide req  to tx 

Explain: 

- Database always start from the beginning and end with commit/rollback

- SQLite no parallel so you could have a deadlock if one transaction is not committed 

- Provide req to link lifecycle transaction to request when the request is succeeded we automatically perform a commit.

Example 2: You have a request then during the request at some point you have database interaction -> then automatically send a begin -< then we immediately  register handlers for the succeeded, failed event of the request

If success we perform a commit

if failed we perform rollback

Then execute SQL

set DEBUG SQLite to start to debug database 

Issue: After the first request to my express handler, my app doesn't respond anymore

First look you see everything is good you provided the req but the problem is it's not find because the req here is express request, not CAP request these are different things

Solution: You can't use them interchangeably so once you're in express, you need to use your own transactions and manage them on your own like this.

Another way for newest version

Async LocalStorage :

Select from the book is actually a cqn object an object representing the query but if you await this then we execute it we'll look at the corresponding transaction to join it to that transaction and it's a bit like magic because we don't provide any rack object here.

So, how does it work how does the CAP framework know which transaction to use which tenant to use, which locate to use? 

The cds is global object so that context is also a global object.But it's not global variable it's local with respect to the async context

Example: Not overwrite the value

Issue: Background database operations have a strange broker

In this example code, you can see that you run backgroundTask then we await a SQL and object. This is a continuation of which transaction to use to join it to. But then you will have nasty race conditions because we use the same transaction and the original request but we don't await the result and that is super nasty. So, this is really problematic and you have race conditions.

 
Solution: using cds spawn. instead of await

Code

• 4. Database pool Configuration

Issue: Some requests fail during high load
There are some potential issues that cause for this problem one course could be for example that you didn't allocate enough memory for your application and in cloud foundry, it's directly correlated to how much CPU you get the more memory you provide the more CPU you get that could be a problem that you run out of memory

Reason: you configure your pool in a wrong way or in a bad way

Solution: Most important options

acquireTimeoutMillis: give you an option to set how long you're allowed to wait to get to acquire a connection out of this pool and if it takes too long throw an error and usually this is set to belittle and that's why some requests get 503

max : is the maximum amount of connections we keep in our pool

Unfortunately, there's no one-size-fits-all solution so you have to test how to adjust these on your application.

• 5. Loggin

Issue: Can't use Kibana to analyze the logs

Normally the log was written like that 

console.log('My log output');

Kibana can't read the log by this way

Solution:

• 6. Generic handlers

Issue: How can i register generic handlers for all services?

Solution: 

create class MyAppService extend our usual application service

Then you can do whatever you want, you can override the init function for example which happens during bootstrap

then you can write a call super init, register some handler that always runs, always perform some log output

• 7. Environment

Issue: How can i easily switch my environment?

Solution: CAP provides profiles which allows you exactly to do it

Then just run by profile

cds env --profile <profile>

cds run --profile <profile>

Example:

db

local: using sqlite

production: using hana

Type cds env to list all profile env

• 8. Testing

Use cds.test for automatically test

How to use:

First create the path to your project, for example you can use a 

path to join current directory  with ".." to go one

Then write a simple test

Example code

• 9. REPL

Using CDS REPL to play around all the CAP API and javascript dynamic language and it has great support for the REPL so we should make use of it.

Just write cds repl

 And now you're in the REPL and you can also do for example you can write const t equals cds.test('./') for this root directory.

And now you can actually interactively deal which running CAP application

await t.GET('/catalog/Books')

Then you will get the response of your Books

You can also connect to the service and get the books

You can also be using raw SQL for select data.

Source: https://www.youtube.com/watch?v=WTOOse-Flj8