Automic Workload Automation

  • Private Community
 View Only

  • 1.  Will deleting 16% of LINUX agents improve response time in v12.3.3 ?

    Posted Sep 16, 2020 11:30 AM
    Hello,

    We have v12.3.3 installed in our test environment.  We have over 3,800 LINUX agents currently active. We have determined that about 600 of these (approx. 16 %) are not needed and can be deleted. We have limited resources in our group so we are trying to determine whether deleting all these agents will improve response time.  We also heard that the Automation Engine can handle up to a million agent connections. Since this is our test environment we want to make sure we will get some response time improvement if we decide to do this task.  We have done a lot of tuning already in this test env. The test env has 20+ clients and we generally have over a hundred users logged on at any given time of the day.

    Has anyone ever done a fairly large-scale removal of agents (over 10 %) ?   And did that result in a decrease in system utilization ?


  • 2.  RE: Will deleting 16% of LINUX agents improve response time in v12.3.3 ?

    Broadcom Employee
    Posted Sep 17, 2020 01:01 PM

    If those Agents aren't doing anything (and operating/connected correctly) the removal should be minimal to impact on utilization.

    Without knowing details, I'd say start by looking at the performance of the db (ae-db connectivity), and looking at when load is placed on the system; is the utilization spiking as a function of heavy load being dropped at one time, or is it all the time?

    If you look at the B.1/10/60 is it all processes or a subset (DWP vs WP, etc)? 


    Original Message


  • 3.  RE: Will deleting 16% of LINUX agents improve response time in v12.3.3 ?

    Posted Sep 18, 2020 12:27 PM
    Hi Darren,

    So we have 2 issues with excessive agents in our test environment. This is the oldest and the least impactful. We have known that we have 500-700 agents (number fluctuates) that aren't being used for a year or more. This is the obsolete agent issue.

    The second more impactful issue is that we also have another 100-200 agents that people spin up as part of Chef or AWS builds. We have seen some response time issues caused by them. This just started this summer and we just started addressing these 100-200 agents this week. We are trying to get the users of these 100-200 agents to stop creating agents that they never need. So hopefully in another 2-3 weeks, we will have this second issue resolved.  Then I will go back to the obsolete agent issue and update this thread.
    Original Message