[22281] When communicating in data sharing mode, if the QoS is set to best-effort, the writer will return an error code RETCODE_OUT_OF_RESOURCES after sending a number of data equal to the history depth. However, if the QoS is set to reliable, this issue does not occur. Why is that?

[Eternity1987]

Is there an already existing issue for this?

• I have searched the existing issues

Expected behavior

When the QoS is set to best effort, data can be continuously sent without any issues.

Current behavior

if the QoS is set to best-effort, the writer will return an error code RETCODE_OUT_OF_RESOURCES after sending a number of data equal to the history depth

Steps to reproduce

a writer and a reader communicate，the data type is bounded and plain，data sharing is ON

Fast DDS version/commit

2.14.3

Platform/Architecture

Ubuntu Focal 20.04 amd64

Transport layer

Data-sharing delivery, Zero copy

Additional context

XML configuration file

No response

Relevant log output

No response

Network traffic capture

No response

[Eternity1987]

Why does the phenomenon occur when the kind of DurabilityQosPolicy is TRANSIENT_LOCAL_DURABILITY_QOS, but messages are sent normally if it is VOLATILE_DURABILITY_QOS? Can you provide some insights into why this might happen, such as not receiving ack or gap

[Eternity1987]

When free_payloads_ is empty, my program doesn't execute advance_till_first_non_removed. A single loan and write process consumes two CacheChange_t in free_payloads_. How can I ensure that release_payload is executed? Otherwise, when a loan has CacheChange_t available, they might not be available by the time write is called.

---

[cferreiragonz]

Hi @Eternity1987. Thank you for reporting this.

We will address it and try to reproduce it in the following weeks.

[cferreiragonz]

Hi @Eternity1987.

Unfortunately, I was not able to reproduce the behavior on Fast DDS v2.14.3. The quickest way to identify the potential issue would be if you could provide some code or a reproducer where this behavior is reflected.

Additionally, it would be helpful if you could share more details about the QoS settings you are using for both the DataWriter and DataReader. In your initial comment, you mentioned that the issue occurs when the ReliabilityQosPolicy is set to BEST_EFFORT_RELIABILITY_QOS, but in a later comment, you suggest it might be resolved by changing the DurabilityQosPolicy to VOLATILE_DURABILITY_QOS. Could you clarify whether the issue is related to one or the other, or if it occurs with a specific combination of both QoS settings?

[Eternity1987]

Hi @Eternity1987.

Unfortunately, I was not able to reproduce the behavior on Fast DDS v2.14.3. The quickest way to identify the potential issue would be if you could provide some code or a reproducer where this behavior is reflected.

Additionally, it would be helpful if you could share more details about the QoS settings you are using for both the DataWriter and DataReader. In your initial comment, you mentioned that the issue occurs when the ReliabilityQosPolicy is set to BEST_EFFORT_RELIABILITY_QOS, but in a later comment, you suggest it might be resolved by changing the DurabilityQosPolicy to VOLATILE_DURABILITY_QOS. Could you clarify whether the issue is related to one or the other, or if it occurs with a specific combination of both QoS settings?

sorry, you maybe close the issue. it's my fault.
I borrowed data and added some information using a struct externally. This struct contains a pointer to the borrowed data, which is then sent. After using it, I call discard_loan. I thought this was fine, and I wouldn't have this problem if I didn't add those extra things externally. Thank you very much. Additionally, could you help me look into this issue? It is critical, and I have also uploaded my project.
#5469

I hope to get a reply. What should I pay attention to in the one-to-many usage of Fast DDS? Is it a problem with my usage or a bug?🤠🤠🤠