TangSiyang2001 commented on issue #17333:
URL: https://github.com/apache/doris/issues/17333#issuecomment-1502702401
My assumption about RPC timeout:
Actually, tasks are offered to work pool with the `try_offer` API now:
```cpp
virtual bool try_offer(WorkFunction func) {
PriorityThreadPool::Task task = {0, func, 0};
return _work_queue.try_put(task);
}
```
rather than the `offer` API:
```cpp
virtual bool offer(WorkFunction func) {
PriorityThreadPool::Task task = {0, func, 0};
return _work_queue.blocking_put(task);
}
```
In currrent code, once the request is processed by the corresponding
bthread, it will be offered to the worker pool immediately and processed in the
async way.
Therefore, RPC timeout under heavy loads may actually result from:
1. (minor) Too few pthreads for bthread workers to manage the requests.
2. (major) The ratio of
`brpc_heavy_work_pool_threads/brpc_heavy_work_pool_max_queue_size` may be
inproper under the corresponding case, so tasks in blocking queue are blocked
too long to be processed by the busy worker threads.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
