Some attachments are not displayed.

[netolyrg]

Describe the bug I am facing the “attachments” problem for Report Portal in all tests: on some runs, the attachments are not displayed and the log is broken into several identical parts (2-8 logs). There are no errors in the server logs.

I don't understand how to debug such a problem. I can gather more information about the environment if someone can help solve this problem.

Steps to Reproduce

1. Launch 2 or more runs of the same test consecutively

Expected behavior All attachments are visible in all launches.

Actual behavior On some runs, the attachments are not displayed.

Package versions Include version info of the following packages: reportportal-client == 5.5.6

Additional context Report Portal 5.11.0 Environment: K8s, storage type: (minio, single), 100+ distributed sync clients.

[HardNorth]

@netolyrg We had such issue on server side in service-api 5.10.0, but this was fixed in 5.10.1. Can you provide any more details?

1. Which test framework do you use? And the agent version? If you are not using any Agent please confirm that you use RPClient class instance (different client implementations use different HTTP clients under the hood).
2. How big is your attachments? It doesn't look big, but just confirm.
3. Did you adjust log_batch_size: int and log_batch_payload_size: int arguments?
4. Do you see any errors in console log, or on service-api side?
5. How was the load on the server when such issues happens?

[netolyrg]

Hi! serviceapi: v5.11.0

1. We use RPClient v5.5.6
2. Not so big, average size < 1MB (png screenshot)
3. Yes, but it did't help and due to our test launcher architecture we can't increase log_batch_size more than 1.
4. No, as far as I know, but I'll check it (as an AQA I don't have access to logs)
5. It happens on all load levels.

[HardNorth]

@netolyrg I'm sorry, you answers are extremely general, I can't nail any problem with them.

[netolyrg]

@HardNorth What additional information I can provide? Logs from service api, exact attachments sizing, load stats? Anything missed?

[rplevka]

logs from the api service would definitely be helpful. re:

the log is broken into several identical parts (2-8 logs).

Are you trying to say that you expected only 1 such log record to be displayed? Can you also check the rabbitmq service logs? How's the service health? isn't it being OOMkilled or something? It appears to me like the workers might process the request but might be killed before they ACK the message?

[HardNorth]

@netolyrg I just found that log_batch_payload_size was supposed to be named as log_batch_payload_limit. My bad, sorry. This issue might results in this bug. Could you please use the latest client version (5.6.7) and check if the issue is still persists?