问题
I have tried really hard to use the embed retryPolicy of GRPC documentation (https://github.com/grpc/proposal/blob/master/A6-client-retries.md#retry-policy) but i fail to understand where i should setup the config in my code.
Ideally i would like the python client to specify its retry policy but i am also interested to understand how to manage it from the server side.
After some digging, i came up with this snipped but it does not work.
import json
from grpc import insecure_channel
service_default_config = {
# see https://github.com/grpc/proposal/blob/master/A6-client-retries.md#retry-policy-capabilities
"retryPolicy": {
"maxAttempts": 5,
"initialBackoff": "1s",
"maxBackoff": "10s",
"backoffMultiplier": 2,
"retryableStatusCodes": [
"RESOURCE_EXHAUSTED",
"UNAVAILABLE"
]
}
}
service_default_config = json.dumps(service_default_config)
options = [
('grpc.service_config', service_default_config)
]
insecure_channel(hostname, options=options)
Can anyone point me out the relevant documentation for me to understand how this works or explain to me what i misunderstand ?
回答1:
I had the same problem with the syntax of the config.
I found this.
In short, the retryPolicy has to be specified as part of a methodConfig which describes how to handle calls to a specific service:
import json
import grpc
json_config = json.dumps(
{
"methodConfig": [
{
"name": [{"service": "<package>.<service>"}],
"retryPolicy": {
"maxAttempts": 5,
"initialBackoff": "0.1s",
"maxBackoff": "10s",
"backoffMultiplier": 2,
"retryableStatusCodes": ["UNAVAILABLE"],
},
}
]
}
)
address = 'localhost:50051'
channel = grpc.insecure_channel(address, options=[("grpc.service_config", json_config)])
where <package>
and <service>
are defined in your proto file.
来源:https://stackoverflow.com/questions/64227270/use-retrypolicy-with-python-grpc-client