Ceph

Environment:
2+1 erasure Model：host Version: Hammer 94.10
Question:
While the PG status is incomplete ，the operation of reading or writing will have a timeout sometimes. I found that every OSD has a resource pool named client_msg_throttler(the maximum is 100) for the client operations. While OSD receives an operation command through the thread of the pipe, it will apply for resource，then it will put the
operation in a queue and wait for processing, while the replicated PG deals with the operation, it will judge whether the PG state is normal. If the PG status is abnormal, the operation will join the waiting queue .In this case, the
resource will not be released untill the PG status is normal. Once it grows to 100,the access to the normal PG will have to hold even if it's the same OSD.
what I want to know is whether there is way to let the abnormal PG release the resource so that the PG in normal status can work.