Hello everyone.
I am looking a way to kill a service when it has exceeded given timeout in seconds.
So basically situation is this: I have created HTTP/JSON API by using Zato’s HTTP channels and services. These services are accessing our PostgreSQL database (SQLAlchemy) and return the result to requesting client.
Generally this setup works fine, but if the request against PostgreSQL takes too long, the following error 504 (Gateway Time-out, The server didn’t respond in time.) is returned.
Main issue here is not that the status 504 is returned, but that the service which was invoked over HTTP channel is never killed in these sitautions.
This consumes server resources for nothing, as the result is never returned to the client (In worser situation, the client could keep on requesting forwever).
For testing solutions and verifying the issue, I’ve created a small service called LoopTest. Running the service takes 100~ seconds to finish and it prints steps while processing.
Services print/debug output is visible for me, as I*m using only one server in foreground mode (zato start server1/ --fg).
from zato.server.service import Service
class LoopTest(Service):
def handle_GET(self):
""" URL: /xxxx/tests/v1/looptest/
"""
for i in range(1,100):
print("Looping i:"+str(i))
time.sleep(1)
So after this long ramble, the actual question is following: What should I do about this?
For now, my only idea is to use a “proxy” service, which will invoke the actual API service with invoke_by_impl_name method (seems to have support for timeouts) or by other means.
Early thanks for any responses.
Ps.
Some people on the forum might suggest optimizing the query and the database and they would be right.
However I’m afraid that there will be similliar issues in the future, which I would not notice until its too late.