Open mtrbpr opened 11 months ago
Hi. I get the same behavior. Even checkhealthy:enabled=false When the RabbitMQ falls and then reruns, Vector stops to send messages into it.
That's hilarious. We are experiencing similar issues with blocking behavior in amqp091-go
client (issue1, issue2) and considered Vector as a replacement. Now it seems I have encountered a similar issue to @sxkote's.
Nov 11 13:07:57 my-rabbit-host.example vector[1234567]: 2024-11-11T13:07:57.516987Z ERROR sink{component_kind="sink" component_id=rabbitmq_sink component_type=amqp}:request{request_id=REQUEST_ID_1}: vector_common::internal_event::service: Service call failed. No retries or retries exhausted. error=Some(AcknowledgementFailed { error: IOError(Os { code: 104, kind: ConnectionReset, message: "Connection reset by peer" }) }) request_id=REQUEST_ID_1 error_type="request_failed" stage="sending" internal_log_rate_limit=true
...
some suppressed logs
...
Nov 11 13:08:13 my-rabbit-host.example vector[1234567]: 2024-11-11T13:08:13.215606Z ERROR sink{component_kind="sink" component_id=rabbitmq_sink component_type=amqp}:request{request_id=REQUEST_ID_2}: vector_common::internal_event::service: Service call failed. No retries or retries exhausted. error=Some(DeliveryFailed { error: InvalidChannelState(Error) }) request_id=REQUEST_ID_2 error_type="request_failed" stage="sending" internal_log_rate_limit=true
This log with InvalidChannelState(Error)
is then logged forever alongside with some suppressed logs, the reconnect does not happen. If restarted, Vector connects and functions as expected.
I suggest this issue should be replicated in (integration) tests. For now I was unable to find out if a root cause is in lapin.
A note for the community
Problem
It seems that disabling sinks healthcheck via setting
--require-healthy
flag tofalse
on startup has no effect when sink type isamqp
and no vhost is specified in amqpconnection_string
, or the vhost does not exist at all.Same thing happens when connection string points to a wrong port:
I tried setting wrong port on other sink types like
kafka
andelasticsearch
and it didn't cause vector to crash so I presume what I'm reporting should be a bug related toamqp
sink or the way vector is using it.Configuration
Version
vector 0.34.1 (x86_64-unknown-linux-gnu 86f1c22 2023-11-16 14:59:10.486846964)
Debug Output
Example Data
No response
Additional Context
vector is installed with apt and rabbitmq is up on port 5672 with following docker-compose deployment:
References
No response