Home page load becomes abnormally slow

[hunghvu]

For some reasons, loading a home page becomes abnormally slow (>5s, even 30s). The server does not seem to have any error. This did not happen a few days ago, and the server has been online since then.

That said, loading an article is still as fast. This leans towards the possibility of MongoDB free tier throttle. Homepage uses dynamic fetching, so it is affected, meanwhile, article page is cached, hence only the first load is slow (but need to validate).

• Try to restart the server.
• See if there is a throttle or bottleneck somewhere (AWS, Cloudflare, MongoDB Atlas).

[hunghvu]

More than 30s from Postman.

[hunghvu]

Indeed the issue is due to MongoDB 10gb bandwidth throttle for M0 tier.
https://www.mongodb.com/docs/atlas/reference/free-shared-limitations/

Because of that, we move to self-hosting the database via Docker for now. It is not mission critical, so we can stand the risk. M2 from MongoDB still has 20 GB limitation per 7 days. Since we are not even sure why 10 GB of M0 is exceeded in the first place, M2 probably will not cut it. The below are steps to migrate from Atlas to Docker host.

1. Install mongodb tool on the local machine (https://www.mongodb.com/try/download/bi-connector, deb package).
2. Perform mongodump --out /path/to/dump --uri="mongodb+srv://username:password". Note, the username and password cannot contains certain character like @, *, &, or else, mongodb tool cannot parse the string.
3. Create a key pair in EC2 Network & security, download the pem file. ED25519 is preferred than RSA.
4. Derive the public key via ssh-keygen -y -f file.pem.
5. Change file permission via sudo chmod 400 file.pem, or else an error due to open permission will show up.
6. Copy and paste the public key to ~/.ssh/authorized_keys on EC2 instance (via AWS console).
7. Copy database dump to EC2 instance via rsync -avzh -e "ssh -i file.pem" /path/to/mongodump/ ec2-user@ip:/path/on/ec2.
8. Install mongodb tool on EC2 instance.
9. Perform mongodb restore via mongorestore --authenticationDatabase=admin mongodb://username:password@host /path/to/mongodump. Keep in mind there should be authenticationDatabase, or else, we see SCRAM-SHA-1 error. Still not sure why, but here is the reference: https://stackoverflow.com/questions/68927857/mongoexport-auth-error-using-mechanism-scram-sha-1

[hunghvu]

Note: the deployment is done, but we will monitor the issue for the next few days. I all go well, we will close this issue afterward.

[hunghvu]

That said, we can keep looking for an alternative managed solution with a better price point.

[hunghvu]

Now utilize the uses of indexes to speed up queries. Also, remove indexes from fields that are underutilized at this stage. In the future, after the search function is implemented, we can update the indexes again.

Theoretically, this should improve performance. Note, indexes creates by Payload are single indexes, not compound indexes.

[hunghvu]

The M0 throttle seems to be lifted now. Anyway, the cluster is upgraded to M2, now we use the self-host container as a backup. The main DB solution is still Atlas.

Consider this issue as resolved now.