Openstatus
www.openstatus.dev
Blog post fly (#419)
* 📝 blog post
* 📝 blog post
authored by
Thibault Le Ouay
and committed by
GitHub
8ba290c4
3bafddf2
+123
apps/web/public/assets/posts/migration-backend-from-vercel-to-fly/fly-dashboard.png
apps/web/public/assets/posts/migration-backend-from-vercel-to-fly/fly-dashboard.png
This is a binary file and will not be displayed.
apps/web/public/assets/posts/migration-backend-from-vercel-to-fly/hono-server.png
apps/web/public/assets/posts/migration-backend-from-vercel-to-fly/hono-server.png
This is a binary file and will not be displayed.
apps/web/public/assets/posts/migration-backend-from-vercel-to-fly/next-server.png
apps/web/public/assets/posts/migration-backend-from-vercel-to-fly/next-server.png
This is a binary file and will not be displayed.
+123
apps/web/src/content/posts/migration-backend-from-vercel-to-fly.mdx
+123
apps/web/src/content/posts/migration-backend-from-vercel-to-fly.mdx
···
1
+
---
2
+
title:
3
+
Why we migrated our backend from Vercel to Fly.io and the challenges we faced.
4
+
description:
5
+
Learn why we switched from Vercel to Fly.io for our backend and the challenges
6
+
we encountered.
7
+
author:
8
+
name: Thibault Le Ouay Ducasse
9
+
url: https://twitter.com/thibaultleouay
10
+
publishedAt: 2023-10-29
11
+
image: /assets/posts/migration-backend-from-vercel-to-fly/fly-dashboard.png
12
+
---
13
+
14
+
In this article we are going to see the reason that made us change our backend
15
+
to Fly.io and the challenges we had during the migrations.
16
+
17
+
We chose [Hono](https://hono.dev/) as our API server with [Bun](https://bun.sh/)
18
+
as the runtime and pick Fly.io as your hosting service.
19
+
20
+
## 🤔 Why did we want to move our backend from Vercel?
21
+
22
+
### ⚡ A lightweight Server
23
+
24
+
We required a lightweight server with a simple REST API for our monitoring
25
+
endpoint. Deploying a simple Express server is possible, but it is not
26
+
specifically designed for this purpose.
27
+
28
+
`It's possible to deploy an Express.js application as a single Serverless Function, but it comes with drawbacks and should only be used as a migration path. Instead, use Next.js or embrace multiple Serverless Functions as you incrementally migrate to the Vercel platform.`
29
+
30
+
[Source Vercel Doc](https://vercel.com/guides/using-express-with-vercel)
31
+
32
+
Also launching a clean and new Next.js server takes 2.5 seconds on my MacBook
33
+
Pro M1 and takes 110mb of RAM, and it includes unnecessary extra features from
34
+
Next.js. Our prod Next.js app takes about 5 seconds to launch on my computer
35
+
(contentlayer).
36
+
37
+
38
+
39
+
For comparaison launching our current server takes 0.19ms and only takes 91mb.
40
+
Our current server stack is Hono + Bun.
41
+
42
+
43
+
44
+
### 💸 Pricing
45
+
46
+
We initially aimed to provide multi-region monitoring for all users while
47
+
maintaining a free tier. On Vercel, if you want a multi-region function, you
48
+
need to opt for Edge Functions. Edge functions are cost-effective as you only
49
+
pay for the actual CPU execution. This means that you won't be billed for idle
50
+
times when fetching data.
51
+
52
+
It's still affordable, but we are a bootstrap business and it difficult for us
53
+
to predict our monthly expenses. If we experience an increase in new users, our
54
+
costs will also increase accordingly.
55
+
56
+
Here's the math for the cost of one monitor for a user:
57
+
58
+
```
59
+
6 (10 min monitors) * 24 * 30 * 3 (average execution unit per monitor) * 6 (number of regions) = 77,760 executions units per month
60
+
61
+
77,600 * (2/1,000,000) = 0.15c per monitor monthly
62
+
```
63
+
64
+
We have over 1000 monitors, and the monthly cost to run them would be $150.
65
+
66
+
While on fly we only have 6 servers with 2vcpu/512Mb It cost us $23.34 monthly
67
+
($3.89\*6).
68
+
69
+
## 🤯 What challenges did we face during our migrations?
70
+
71
+
### Docker + monorepo = 🪨
72
+
73
+
We are deploying to fly.io. We have to setup our app as a Docker image. Our apps
74
+
is in a monorepo. Our initial image was over 2 GB in size, which was excessively
75
+
large for a basic server.
76
+
77
+
Our image included everything, which was inefficient. After optimizing, our
78
+
image now occupies only 700MB. Although it is still somewhat large, it is a
79
+
significant improvement over our initial version.
80
+
81
+
It was something we never had to manage with Vercel deployment.
82
+
83
+
### ⏳ Fly deployment timed
84
+
85
+
Our Fly deployments have been experiencing frequent timeouts without any
86
+
specific reason. The only solution we have discovered is to increase the timeout
87
+
duration.
88
+
89
+
```
90
+
flyctl deploy --wait-timeout=500
91
+
```
92
+
93
+
Based on our experience, Fly deployments are generally less reliable compared
94
+
(more timed out) to Vercel deployments. Additionally, we have not discovered a
95
+
quick method to rollback to the previous version.
96
+
97
+
### 🐛 The Bun bug
98
+
99
+
When we migrated to Fly, we decided to use Bun as our runtime. However, in the
100
+
first few hours after the migration, we observed an unexplained increase in
101
+
request failures.
102
+
103
+
After digging into the Bun GitHub we found a solution: We needed to set the
104
+
`keepalive` parameter to `false`. This is necessary because closed connections
105
+
are not automatically removed and remain in the `CLOSE_WAIT` state.
106
+
107
+
Here's the GitHub issue:
108
+
109
+
https://github.com/oven-sh/bun/issues/3327
110
+
111
+
I wish it had been documented elsewhere.
112
+
113
+
## Our conclusion
114
+
115
+
We are pleased with our migration to Fly.io, although it was accompanied by a
116
+
challenging weekend. And we still love Vercel, they offer a super product, it
117
+
removes a lot of pain for the developers.However, if you require a hosting an
118
+
application other than Next.js, it may not be the best option.
119
+
120
+
We are still using it for our frontend.
121
+
122
+
Create an account on [OpenStatus](https://www.openstatus.dev/sign-up?ref=blog3)
123
+
to start monitoring our website and managing your incidents for free.