query regarding deployment

[#1]

Hello All,

Please help me to find the answer of the query mentioned below:-

At the time of load testing we stop few MPs and the same time if someone do deployment of proxied will that deployment will be done on that MP or deployment will be done on another MPs leaving the unreachable MP.

Note:-

We will not shutdown the MP but through curl command we will make it unreachable, so that no traffice will be routed to it for processing.

Thanks.

[dchiesa1 · Accepted solution]

Your question is not clear.

and the same time if someone do deployment of proxied will that deployment will be done on that MP or deployment will be done on another MPs leaving the unreachable MP.

That doesn't parse. But I'll provide some information and maybe it will answer your question.

From your question I'm assuming you're using Apigee Edge OPDK.

If you deploy something (an apiproxy, a sharedflow) to "an environment", then it will be marked as deployed. If there are MPs that are part of "the environment" that are not running and available, then the deployment does not reach these MPs. If you restart the MPs that were previously unavailable, the deployment will propagate to these MPs. If you add MPs to the environment, the deployment will propagate to those MPs too.

You can verify all this by testing it yourself.

If you simply make the MPs unreachable, they still receive deployments. They just don't handle incoming API requests. After you mark them reachable again, they will begin handling inbound API requests.

Test it for yourself.

[#3]

Hello Dino,

Thanks for the reply and sorry if the question was quite confusing. But your answer cleared my doubt.

Thanks a lot.

[balajoshi1986]

Hello Dino,

One thing I observed is that when I set the reachable flag to false through curl command from remote server and after that when I check the status from the same server it shows reachable as false but when I go to that message processor and run the curl it shows the reachable true. Also I am using 4.18.01 version of apigee and as per their documentation they are using apigee-service command to set the message processor as unreachable.

Does curl command work on 4.18.01 version and if above statement is correct than is it that router first contact MS server to check the MPs reachability flag and then directs the traffic.

Thanks.

[dchiesa1]

Hi - I don't know why you'd be seeing that behavior. The apigee-service command will just rely on the administrative APIs that turn reachability on and off. Curl should also work on 4.18.01.

But the real question is: does the MP serve traffic when it is marked not-reachable using the apigee-service command? If it is correctly not serving traffic, then you should be good to go, right? Irrespective of the value you get back from curl on the local MP.

If that inconsistency bothers you, I suggest that you contact Apigee support to ask for help in troubleshooting it and diagnosing it.

[#6]

Hello,

Thanks for the reply.

If the reachable flag is off still traffic is flowing and giving 200 status code on the trace session only if "isUp" flag is flase than we are getting 502 status code. We have raised the issue with apigee support waiting for thier answer.

The flow should be :-

router confirms the reachable MPs from MS server and direct the traffic to the MPs.

Or,

router check the port opened on MP and directly sends the traffic.

I observed that when reachable flag is off still than port 4528 was opened on the server and MP was accepting the requests but when isUp flag was off the port 4528 was available and traffic was not flowing.

Thanks.

[balajoshi1986]

Hello Dino,

After adding below properties it seems that router is now acknowledging MPs reachable flag.

conf_load_balancing_load.balancing.driver.nginx.target.pool.check=interval=5000 rise=1 fall=2 timeout=3000 type=http port=8082 default_down=true conf_load_balancing_load.balancing.driver.nginx.target.pool.check_http_send="GET /v1/servers/self/reachable HTTP/1.0\\\\r\\\\nConnection: keep-alive\\\\r\\\\n\\\\r\\\\n"

Thanks.

[dchiesa1]

Great! Sounds like it's resolved.

[#9]

Yes Sir, Thanks for your efforts.