https://github.com/lewiscowles1986/experiment-aws-lambda-lifetime-and-caching-python
Just a sample experiment I did, based on a talk with a colleague
https://github.com/lewiscowles1986/experiment-aws-lambda-lifetime-and-caching-python
Last synced: 3 months ago
JSON representation
Just a sample experiment I did, based on a talk with a colleague
- Host: GitHub
- URL: https://github.com/lewiscowles1986/experiment-aws-lambda-lifetime-and-caching-python
- Owner: Lewiscowles1986
- Created: 2024-08-03T12:03:16.000Z (10 months ago)
- Default Branch: main
- Last Pushed: 2024-08-04T06:17:04.000Z (10 months ago)
- Last Synced: 2025-03-06T14:55:04.194Z (3 months ago)
- Language: Python
- Size: 24.4 KB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# AWS Lambda experiment, Python cachetools
This is an experiment aimed at showcasing cachetools in AWS Lambda
It comes from a conversation with Engineer Megan List.
I Asserted that AWS Lambda can:
- [X] cache values for more than 15 minute maximum runtime
- [X] take memory snapshots of VM to get faster warm boot times
- [ ] I was wrong when I said it reaps warm lambdas over 1 hour old.
To check this, I'm monitoring the PID, as well as the time, and trying to call serially, avoiding parallel calls.
I have no dependencies on SSM or anything else, because this is a simple app. YMMV as we do not control the scheduler, re-balancing of machines etc. That is part of what is good about serverless.
As stated above, I was wrong about lambdas over 1 hour old being reaped. Although the second set of logs (where I remembered to use `-r` flag, to actually recurse the folder) shows this. Perhaps this is part of AWS coordination, which in a prior test run had looked like it happened. Honestly I cannot remember if it's a day, or they just run forever; but it's definitely not within the control of the user; and I've seen both ~1 hour and over that.
## Pre-requisites
- Curl 8.6+
- Terraform 1.5+
- cli `zip` command compatible with:
- `zip `
- `zip -ur ../ ./*`
## Usage
1. Initialize terraform and apply
```
terraform init
terraform apply
```
2. Record the URL output. Mine and the one given below was https://msxbosdarhxhgoxzcyeo2xjtdu0yhkzz.lambda-url.us-east-1.on.aws/
3. Run the test
```
for i in {1..60}
do
curl https://msxbosdarhxhgoxzcyeo2xjtdu0yhkzz.lambda-url.us-east-1.on.aws/ > logs/$(date +"%Y-%m-%dT%H%M%S").json
sleep 60
done
```
## Sample output
I've saved the output of the logs in the logs-example.zip by using `zip -r logs-example.zip logs/` in the hopes it might help budget strained folks interested in the output.
Note that the sample output was manual commands, until it became the loop example above at 11:33:00
The loop iterated a few times, which is why you see it go back to 00 in seconds. There is no magic there, just fast fingers, and restarting with ``+``, then `` then ``.
## Further thoughts
I wonder if I could exploit this property for other cool demo's. Like fan-out, fan-in data using Lambda, or sequential flows in a single lambda-lith.