Hi, I am executing the following REST API call to start all the services on a given node. I see two different behaviors. When I run it from the command line, it runs almost instantaneously and I see that all the services are started. However when I call the very same API from a shell script which is in turn called from our framework, it takes nearly 10m to start all the services. I see a few of the services saying "INSTALLING" for quite some time before they start up.
are you sure your framework is not adding latency to the API calls? One you rule that out, you should see if Ambari heapsize is set adequately. Also please help us by answering the following: how many nodes are you running, what version of HDP is it and why are you still using bari 2.1? We recently identified an issue with Ambari prior to 188.8.131.52 that take a long time to execute on large clusters, though I'm leaning towards issues woth your framework and/or Ambari has sufficient resources.
Artem, I am running it on a test setup of 2 nodes with 12 instances spread across these nodes. The ambari server instance has 1GB of RAM and 1 core allocated to it which I presume is not enough. However even the cli call was issued to this very same instance so what could be the difference. There two servers are quite beefy with about 137GB of RAM in total.
That's why I asked how sure are you your framework does not introduce any latency. Please follow this guide to increase Ambari heapsize https://community.hortonworks.com/questions/622/how-to-increase-the-ambari-heap-size.html