Do not consume rate limit token on invalid tasks

[dhiaayachi]

When setting maxTaskQueueActivitiesPerSecond to enable rate limit on task queue, the token is consumed when task is attempted to be delivered. However, the task could be invalid (expired or workflow already closed), which lead to waste of rate limit token.

We should change this to only consume token when task is delivered.

[dhiaayachi]

Thank you for reporting this issue! I understand your concern about the rate limit token being consumed when a task is attempted to be delivered, even if it's invalid.

Currently, Temporal's rate limiting mechanism consumes a token when a task is attempted to be delivered, regardless of its validity. This behavior is documented in the Rate Limiting section of the Temporal documentation.

We are aware of this potential for wasted tokens and are actively working on a solution that will consume tokens only when tasks are successfully delivered.

In the meantime, you can consider using a workaround such as:

• Setting a lower maxTaskQueueActivitiesPerSecond value: This can help to reduce the number of wasted tokens.
• Implementing custom logic in your workflows: You could implement custom logic within your workflows to handle invalid tasks and prevent them from being attempted in the first place.

We appreciate your feedback and will keep you updated on the progress of this issue.

[dhiaayachi]

Thank you for reporting this issue!

We understand that the rate limit token is currently consumed when a task is attempted to be delivered, even if the task is invalid. This can lead to wasted tokens.

We are currently working on improving this behavior by only consuming tokens when the task is successfully delivered.

In the meantime, you can consider the following workarounds:

• Increase maxTaskQueueActivitiesPerSecond: This will give you more tokens to work with, reducing the impact of wasted tokens.
• Use a custom task queue handler: You can implement your own task queue handler that only consumes tokens when a task is successfully delivered. This will require more development effort, but will give you more control over how rate limiting is applied.

We will update you on the progress of the fix.

[dhiaayachi]

Thank you for reporting this issue! I understand that the rate limit token is currently being consumed when a task is attempted to be delivered, even if the task is invalid. This can lead to wasted rate limit tokens.

We are currently investigating this issue and working on a solution. In the meantime, you can work around this by:

• Setting the rate limit to a higher value: This will ensure that you have enough tokens to handle any invalid tasks.
• Implementing your own rate limiting mechanism: You can use a custom rate limiter to control the rate at which tasks are attempted to be delivered.

We apologize for any inconvenience this may cause. We will keep you updated on the progress of this issue.