Top 3 pain points for serverless developers
ฝัง
- เผยแพร่เมื่อ 21 ก.ค. 2024
- This is the 100th episode of Serverless Expeditions! To celebrate, Martin and Wes discuss 3 top obstacles for serverless developers and how to overcome them. Watch along to learn how an old circuit board and imaginary maps can help you think differently about serverless computing.
Chapters:
0:00 - Intro
0:28 - First obstacle: user authentication
2:46 - Second obstacle: database connections
4:33 - Martin’s app without a database
6:20 - Third obstacle: technical architecture
8:00 - Think differently about serverless computing
9:31 - Wrap up
Permissions on Google Cloud Run → goo.gle/3EOzuoo
How to authenticate calls to your Google Cloud Run service → goo.gle/3MJ45W9
Securing a REST API with JWT, concepts → goo.gle/3ERQ7zl
Securing a REST API with JWT, code → goo.gle/3DhE8u5
Centralize access to your organization’s websites with Identity Aware Proxy (IAP) → goo.gle/3TyvoFa
Creating a REST API with Node.js and MySQL → goo.gle/3SkK9dK
Hosting a REST API with a Cloud Firestore backend → goo.gle/3EOJJZJ
Cloud Run + Cloud SQL, setup → goo.gle/3g7SmEs
Optimize application performance with Cloud Run and Cloud SQL → goo.gle/3gk0cec
Building a highly available Cloud Run app with Cloud Spanner → goo.gle/3MIgJVp
Designing a serverless finance system on Google Cloud → goo.gle/3VDKmLJ
Designing a serverless app on Google Cloud → goo.gle/3VGaIwP
How to port stateful web apps to Cloud Run → goo.gle/3Tb81BN
Watch more episodes of Serverless Expeditions → goo.gle/ServerlessExpeditions
Subscribe to Google Cloud Tech → goo.gle/GoogleCloudTech
#ServerlessExpeditions - วิทยาศาสตร์และเทคโนโลยี
![รวมเพลงลูกทุ่ง 6 หนุ่มไทบ้าน ฮิตติดกระแส ชุดที่ 1 l มนต์แคน ไมค์ ไผ่ เสถียร ไหมไทย พี [Longplay]](http://i.ytimg.com/vi/GD2v1OjMw5I/mqdefault.jpg)
💯 episodes and counting of Serverless Expeditions and it all thanks to you - our developer community. Thank you!
🔔 Subscribe so you never miss a serverless expedition → goo.gle/ServerlessExpeditions
So Nice the dynamic and how in the final you bring the convention ideas to explain, really great
Exactly the opposite.
8:47 exactly whatever on-prem they copy to the cloud, i came to a company they copy paste their on-prem k8s to gke, i'm changing it using cloud build, skaffold, cloud deploy, cloud run for some projects ...
Wow, copy-paste deployments sound pretty risky. I'm glad to hear you're fixing that!
Thanks for another great video! Your videoes are amazing for getting a fast and unintimidating understanding of GCP. 🤩
At 3:22 you mentioned that relational databases are good for aggregating data over many records, and then you mentioned MySQL, PostgreSQL, and SQL Server.
Those databases are notoriously bad for doing aggregations over many records. A report here and there may be ok, but on a production database, you would avoid mixing CRUD operations and multi-row aggregations in the same database.
If multi-row aggregations are essential for your app, consider a column store database instead. GCP offers both BigQuery and Snowflake out of the box.
Thank you for the kind words, Data Maverick!
You bring up a good point about data aggregation. I think it's a matter of degrees. Data aggregation can be done in multiple ways. From slowest to fastest:
1. NoSQL database.
2. SQL database.
3. Analytics database, like BigQuery or Snowflake.
In the video we were talking about going from method 1 to 2. You're saying that method 3 is even faster, which I agree with. For some applications, method 2 is good enough. Others require method 3.
@@TheMomander You are 500% right!!!
Integrating an analytical database into an app takes considerable effort and upkeep.
So 49 out of 50 times, it is better to "just" use the best database for the app and then take the performance hit doing aggregation.
Have you covered already how to implement a fully serverless data platform (or lakehouse) on GCP?
I want to watch in an upcoming video, what Serverless option to take when going Serverless? what Serverless options is more suitable for some kind of application?
Excellent suggestion, thank you! I'm adding that to my list.
I find it difficult to maintain a consistent redis connection with cloud run. Socket disconnects all the time. We tried heart beats as well. It's failing on cold start and also when we have minimum instances. So we can build a consistent application on cloud run. Is there a recipe on how the redis connection should be handled?
That's interesting; I haven't heard of this problem before. Is the Redis instance in Google Cloud or elsewhere? Do you use "Memorystore for Redis" or have you installed Redis on a virtual machine yourself?
@@TheMomander Thanks for responding. We are using "memory store for redis" solution with Cloud run. Cloud run and Memory Store for Redis are connected to the same VPC. They connect and operate without any issues. Like I said, even heart beats are okay. But at a certain time for some strange reason the connection is lost. So all subsequent commands fail. We often joke that the issue happens when we are planning to do some demo to a client. There is no particular pattern on when it happens. Once a week or twice a week. The issue was more frequent when we had minimum instances running than when the cloud run scales to zero. I haven't found much on this issue online so was wondering if it had something to do with some configuration on VPC or redis client. We are using dart server at the moment but is planning to move to express.js in the hope that a more frequently maintained npm redis client library would give us better performance.
@@vineethp8168 Thank you for writing this up. I agree that it would be a good idea to switch to a more commonly used runtime. Please add another comment when you've switched and let me know if it helped or not!
@@TheMomander we started using node JS package - redis 4.1.0 with MemoryStore for Redis. To test against what we already had, we had two instances of Memorystore running - one with the older dart Redis client and the other with the new Node J's redis client. Both of them have been running without the issue for over a month now since I messaged you the last time and then failed exactly the same time today. Which makes me wonder if it has something to do with some time bound process. How do other GCP users handle this scenario?
@@vineethp8168 Thank you for the update. I've asked around and I have not heard anything about this. Sorry I couldn't be of more help.
I love cloud run, but attaching a redis memory store on basic plan costs at least 50 € / month. Attaching a dedicated volume using Filestore is way to pricy with 200 € / month.
Why not provide smaller and cheaper possibilities to attach volumes. Especially smaller apps would benefit a lot from cheap persistent volumes.
Yes, there's a bit of a cost for using Cloud Memorystore for Redis. Awhile back, I produced content to help App Engine users migrate from its Memcache service to Memorystore/Redis where I give hints on minimizing costs. Search for "Migrating from App Engine Memcache to Cloud Memorystore" to find the post, video, and tutorial.
As far as Cloud Filestore goes, do you absolutely need a live NAS/NFS-mounted filesystem attached to it? What is your use case? Another alternative to your architecture is to configure a vanilla Compute Engine VM running Redis with an attached persistent disk (much cheaper than Firestore and mountable on other VMs) and firewall. Yes, that's not a "serverless" solution. The team is looking into some kind of persistent volume for serverless platforms, but that's just an idea for now, hence why I asked about your use case.
sir plz explain what is dry run ?
It can mean many things. In what context have you heard "dry run" mentioned? I don't believe we used this term it in the video.
Cloud Functions for Firebase need support for Python, Java & Go at minimum. Stop forcing Typescript /Javascript on the backend
If you want to write functions in Python, Java, Go, C#, Ruby, or PHP, try out the regular Cloud Functions (without Firebase). You won't be able to use the "firebase deploy" command, but you will have more language choice.