Decreasing write capacity was disallowed without any reasons, decreasing using Management Console worked about dynamodb-lambda-autoscale

[tmitchel2]

Hey, yes you're completely right, it is a bug.

I actually noticed that recently and have fixed it on an experimental branch where i am performing a large refactor. https://github.com/channl/dynamodb-lambda-autoscale/blob/experimental/src/ProvisionedThroughtputCalculator.js You can see my change here. Unfortunately just at the moment I have zero spare time to look into this but happy to accept a simple pull request. Otherwise i will soon be merging my refactor to main once fully tested.

[chris4826]

Hi,
thank's for your feedback.
I compared the files and tried to find a solution.
Unfortunately I have to use your commit dccd224 because the newer commits do not work for me. You have reduced the number of packages in packages.json. When running "npm run build" I'm getting a lot of error, e.g. "gulp dist" not found. Do you already know these problems, too?

So I tried to fix the issue with the old commit where src/utils/ThroughputUtils.js doesn't exists. But src/utils/Throughput.js consists of similar code.

I introduced a new variable for writes:
let isWriteDecrementAllowed = adjustmentContext.AdjustmentType === 'decrement' ? RateLimitedDecrement.isDecrementAllowed(data, adjustmentContext, d => this.calculateDecrementedWriteCapacityValue(d)) : true;

And then changed isAdjustmentAllowed with an addtional or operator:
let isAdjustmentAllowed = isAfterLastDecreaseGracePeriod && isAfterLastIncreaseGracePeriod && (isReadDecrementAllowed || isWriteDecrementAllowed);

I will let you know if this small fix works properly after doing some load test. If you already know that this won't work please let me know.

[tmitchel2]

Yea that doesn't look quite right to me because you would allow a decrement on read if either read or write decrement is allowed, and then the same for writes.

[chris4826]

I agree with you. I will probably wait till you merge the experimental branch into master because your refactoring is too big in order to select only specific piece of code from the experimental branch.