Additional CPU burst capacity (at launch) is missing from new Lightsail instances
I'm currently testing Lightsail to determine whether or not it would be feasible to migrate my current fleet of game servers from a competing cloud provider. The Lightsail documentation states that instances are started with launch CPU burst capacity ("launch mode"), to help ensure that you have available burst credits for your launch scripts:
https://docs.aws.amazon.com/lightsail/latest/userguide/cpu-burst-capacity-accrual.html
Lightsail instances receive additional CPU burst capacity at launch, this is called launch CPU burst capacity. Launch CPU burst capacity allows instances to burst immediately after launch before they have accrued additional burst capacity. Launch CPU burst capacity does not count towards the burst capacity limit. If your instance has not spent its launch CPU burst capacity, and remains idle over a 24-hour period while accruing more burst capacity, its CPU burst capacity (percentage) metric graph will appear as over 100%.
Additionally, some Lightsail instances start in launch mode, which temporarily removes some of the performance limitations that are typically present on burstable instances. Launch mode allows you to run resource-intensive scripts at launch without affecting the overall performance of your instance.
However, this does not appear to be happening. Here are two re:Post questions stating the same issue, without a real answer:
https://repost.aws/questions/QUWDUliXC0RE-OiFjZOmiY4w/lightsail-instances-has-there-been-a-change-to-launch-mode-burst-capacity-for-new-instances - Notice the screenshot mentioned in the comments (https://snipboard.io/mNuO2r.jpg) which shows the documented behavior that should be occurring.
I've attached 2 images. The first image (Example 1) shows an example of my instance that was deployed with my launch script:
The second image (Example 2) shows an example of my instance that was deployed without my launch script:
Notice that the initial burst capacity starts at 0% in both cases, which differs from the statements in the first quoted paragraph above. Additionally, the instance that was deployed with my launch script (Example 1) clearly shows throttling at 20% CPU utilization, which differs from the statements in the second quoted paragraph above (though, that second paragraph says "some" instances, not "all"; I'm curious where this applies).
These instances were provisioned via Terraform using the small_3_0 bundle and the alma_linux_9 blueprint. I've also tested with amazon_linux_2023.
So my question is, is the documentation wrong, or is this not working as intended? And if it is not working as intended, how do I inform AWS of this without purchasing a support plan? I am still evaluating whether or not it would be a good choice to migrate to AWS, and I'm not going to spend money on support just to ask them if their documentation is wrong.
- Newest
- Most votes
- Most comments
For anyone who's having the same issue, support finally got back to me (3 weeks after I opened the ticket!!) with the following:
In your case the "launch mode" does not apply for the combination of bundle as "small_3_0" and blueprint as "alma_linux_9". They have also informed that they will be making the required changes and will clarify the documentation.
In other words, the documentation is lacking critical information.
Based on the information you've provided and the sources available, it appears that there may be a discrepancy between the documented behavior and what you're experiencing with your Lightsail instances. The documentation indeed states that Lightsail instances should receive additional CPU burst capacity at launch, allowing them to burst immediately before accruing additional burst capacity.
However, your observations and the examples you've shared indicate that this launch CPU burst capacity is not being applied as expected. Both of your instances start with 0% burst capacity, which contradicts the documentation's statement about instances receiving additional burst capacity at launch.
The throttling you're seeing at 20% CPU utilization in Example 1 (with your launch script) also seems to conflict with the documentation's description of "launch mode," which should temporarily remove some performance limitations for resource-intensive scripts at launch.
Given that you've tested this with different Linux distributions (Alma Linux 9 and Amazon Linux 2023) and are seeing consistent behavior that doesn't match the documentation, it's possible that there might be an issue with the current implementation of launch CPU burst capacity in Lightsail.
As for informing AWS about this potential discrepancy without purchasing a support plan, you have a few options:
-
You can post your findings on the AWS re:Post forum, providing detailed information about your setup and observations. This might attract attention from AWS staff or community experts who could escalate the issue internally.
-
You could try reaching out to AWS through their general customer service channels or social media platforms to report the potential documentation inaccuracy or service issue.
-
If you're part of any AWS user groups or attend AWS events, you might be able to bring this to the attention of AWS representatives there.
It's important to note that while these methods might help bring attention to the issue, they may not guarantee a quick resolution or detailed explanation without a support plan. However, raising awareness about the discrepancy could potentially lead to clarification or updates to the documentation if there have been changes to how launch CPU burst capacity is implemented in Lightsail.
Sources
View CPU burst capacity accrual for Lightsail instances - Amazon Lightsail
Instances - Amazon Lightsail
Detect Lightsail instance bursting for optimal performance - Amazon Lightsail
Relevant content
- AWS OFFICIALUpdated a year ago
- AWS OFFICIALUpdated 2 years ago
