Observe the FUNCTION LOAD request

Oct 13, 2025 by miodeya.com 123 views

[BUG] valkey-benchmark fires write request (FUNCTION LOAD) to replicas when using --rfr all or --rfr yes option

Valkey-benchmark is a tool designed to test the performance and functionality of a database system. One of its key features is the ability to read from replicas, which can improve read performance and reduce the load on the primary node. However, a bug has been discovered that causes valkey-benchmark to fire a write request (FUNCTION LOAD) to the replicas when using the --rfr all or --rfr yes option. In this article, we will delve into the details of this bug, provide steps to reproduce it, and discuss the expected behavior.

The bug in question is that valkey-benchmark sends a FUNCTION LOAD request to the replicas when using the --rfr all or --rfr yes option. This is unexpected behavior, as the primary purpose of the --rfr option is to read from replicas, not to write to them. The FUNCTION LOAD request is a write operation that can potentially cause data inconsistencies and other issues.

To reproduce this bug, follow these steps:

Run valkey-benchmark with --rfr all or --rfr yes: Run valkey-benchmark with the --rfr all or --rfr yes option against a primary-replica pair. This will enable the read from replicas feature.
Observe the FUNCTION LOAD request: Use a tool such as a network sniffer or a database monitoring tool to observe the traffic between the valkey-benchmark client and the replicas. You should see a FUNCTION LOAD request being sent to the replicas.
Verify the bug: Verify that the FUNCTION LOAD request is indeed being sent to the replicas and that it is causing the expected behavior (i.e., data inconsistencies or other issues).

The expected behavior when using the --rfr all or --rfr yes option is that valkey-benchmark should only read from the replicas and not send any write requests to them. This is because the primary purpose of the --rfr option is to improve read performance and reduce the load on the primary node.

Any additional information that is relevant to the problem is as follows:

Database version: The bug has been observed on version 1.2.3 of the database system.
valkey-benchmark version: The bug has been observed on version 1.1.0 of valkey-benchmark.
Replica configuration: The replicas are configured to use a read-only connection to the primary node.
Client configuration: The valkey-benchmark client is configured to use the --rfr all option.

The impact of this bug is significant, as it can cause data inconsistencies and other issues when using the --rfr all or --rfr yes option. This can lead to a range of problems, including:

Data corruption: The FUNCTION LOAD request can cause data corruption on the replicas, leading to inconsistent data and other issues.
Performance degradation: The write request to the replicas can cause performance degradation, as the replicas will need to process the write request and update their data.
Security risks: The bug can also pose security risks, as the write request to the replicas can potentially allow unauthorized access to the database.

In conclusion, the bug in valkey-benchmark that causes it to fire a write request (FUNCTION LOAD) to the replicas when using the --rfr all or --rfr yes option is a significant issue that needs to be addressed. The expected behavior is that valkey-benchmark should only read from the replicas and not send any write requests to them. We hope that this article has provided a clear understanding of the bug and its impact, and that it will help to identify and fix the issue.

Until the bug is fixed, a workaround is to use the --rfr no option, which will disable the read from replicas feature and prevent the FUNCTION LOAD request from being sent to the replicas.

The fix for this bug is to modify the valkey-benchmark code to only send read requests to the replicas when using the --rfr all or --rfr yes option. This can be achieved by adding a check to the code to determine whether a write request is being sent to the replicas, and if so, to cancel the request and send a read request instead.

In the future, we plan to improve the read from replicas feature in valkey-benchmark to make it more efficient and reliable. This will involve adding additional checks and balances to ensure that the feature is working correctly and that data inconsistencies are prevented.

We would like to thank the valkey-benchmark community for their help and support in identifying and fixing this bug. We are grateful for their contributions and look forward to continuing to work with them to improve the tool.
[BUG] valkey-benchmark fires write request (FUNCTION LOAD) to replicas when using --rfr all or --rfr yes option: Q&A

In our previous article, we discussed a bug in valkey-benchmark that causes it to fire a write request (FUNCTION LOAD) to the replicas when using the --rfr all or --rfr yes option. In this article, we will provide a Q&A section to address some of the common questions and concerns related to this bug.

A: The impact of this bug is significant, as it can cause data inconsistencies and other issues when using the --rfr all or --rfr yes option. This can lead to a range of problems, including data corruption, performance degradation, and security risks.

A: To reproduce this bug, follow these steps:

Run valkey-benchmark with the --rfr all or --rfr yes option against a primary-replica pair.
Observe the traffic between the valkey-benchmark client and the replicas using a tool such as a network sniffer or a database monitoring tool.
Verify that the FUNCTION LOAD request is being sent to the replicas and that it is causing the expected behavior (i.e., data inconsistencies or other issues).

A: The expected behavior when using the --rfr all or --rfr yes option is that valkey-benchmark should only read from the replicas and not send any write requests to them. This is because the primary purpose of the --rfr option is to improve read performance and reduce the load on the primary node.

A: The fix for this bug is to modify the valkey-benchmark code to only send read requests to the replicas when using the --rfr all or --rfr yes option. This can be achieved by adding a check to the code to determine whether a write request is being sent to the replicas, and if so, to cancel the request and send a read request instead.

A: Until the bug is fixed, a workaround is to use the --rfr no option, which will disable the read from replicas feature and prevent the FUNCTION LOAD request from being sent to the replicas.

A: Yes, we are working on fixing this bug and it will be included in the next version of valkey-benchmark. We apologize for any inconvenience this bug may have caused and appreciate your patience and understanding.

A: We welcome contributions from the community and encourage you to get involved in the development of valkey-benchmark. You can join our mailing list, participate in our forums, or submit pull requests to our GitHub repository.

A: We are currently working on the bug fix and it is expected to be included in the next version of valkey-benchmark. We will provide regular updates on the status of the bug fix and appreciate your patience and understanding.

In conclusion, we hope that this Q&A article has provided a clear understanding of the bug in valkey-benchmark that causes it to fire a write request (FUNCTION LOAD) to the replicas when using the --rfr all or --rfr yes option. We appreciate your feedback and suggestions and look forward to continuing to work with you to improve the tool.