juzhiyuan commented on issue #12524:
URL: https://github.com/apache/apisix/issues/12524#issuecomment-4366374755
In APISIX 3.13, the skywalking tracer pushes finished trace segments into
ngx.shared.tracing_buffer. The skywalking-logger endpoint_addr only configures
log reporting to /v3/logs; it does not configure where the skywalking tracer
sends /v3/segments.
From the screenshots, I can see tracing_buffer is already set to 500m, but I
do not see the config.yaml plugin_attr.skywalking section. Please confirm
whether you configured:
```yaml
plugin_attr:
skywalking:
report_interval: 3
endpoint_addr: http://<your-oap>:12800
service_name: ...
service_instance_name: ...
```
The route-level skywalking plugin config should mainly contain sample_ratio.
The report_interval and endpoint_addr for trace segment reporting should be in
plugin_attr.skywalking.
If _plugin_attr.skywalking.endpoint_addr_ is not set correctly, APISIX may
still try to report trace segments to the default http://127.0.0.1:12800. Under
_ab -c 2000 -t 20 -k_ with _sample_ratio=1_, trace segments can be produced
much faster than they are drained, and tracing_buffer can eventually return no
memory.
Could you also check whether error.log contains Segment report fails,
Instance report fails, or Agent ping fails during the test? It would help to
test these three cases separately:
1. only skywalking enabled
2. only skywalking-logger enabled
3. both enabled
For skywalking-logger, batch_max_size=10000 plus
timeout=100/max_retry_count=5 can also create large pending batches if OAP is
slow or unreachable, but that is separate from the tracer.lua:145 error path.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
