Join the Customer Advisory Board

WebSphere & Liberty CAB

Join this online group to communicate across IBM product users and experts by sharing advice and best practices with peers and staying up to date regarding product enhancements.

 View Only
  • 1.  WAS metrics – Which enhancements would you like to see most?

    Posted Fri March 19, 2021 11:02 AM
    We'd like to hear back from you. If you did not join the WAS CAB session today, you can catch up here

    Which enhancements would you like to see most? What else are we missing?

    Improve metrics labelling

    -Add cell label to metrics
    -Add cluster label to metrics
    -Add appname label to EJB metrics

    More flexible query options

    -Provide ability to omit nodeagent metrics from results


    Improve resiliency in cells with very slow servers

    -Add timeout for server metrics scraping
    -Refresh node/server list asynchronously
     

    Improve support for Kubernetes deployments

    -Provide property to enable/disable cell/node/server labels on WAS Prometheus output
    -Publish a Grafana dashboard with filtering options for namespaces / pods / containers

    Improve observability

    -Include server scrape duration metrics in metrics endpoint output


    ------------------------------
    Claudia Beisiegel
    claudiab@us.ibm.com
    Join our CAB sessions every Thursday and Friday at 9:15 am EST
    ------------------------------


  • 2.  RE: WAS metrics – Which enhancements would you like to see most?

    Posted Thu August 26, 2021 03:58 PM
    Please may I first thank this WAS CAB for providing this metrics application. We have deployed this across all of our mainframe WAS servers and we are thrilled with the level of detail. I understand that what can be collected is determined on what is enabled with regards to PMI. That said we would really like to see the Native Memory used by a server externalized via a PMI metric. Currently we are running a Display PID every so often for our mainframe servers, parse the output and insert into a DB2 table. Not optimal and certainly not visible to Grafana.  Thanks Steve Baker, Fidelity Investments

    ------------------------------
    Steve Baker
    Director System Programming
    Fidelity Investments
    Chapel Hill NC
    9194581934
    ------------------------------



  • 3.  RE: WAS metrics – Which enhancements would you like to see most?

    Posted Sun August 29, 2021 06:54 PM
    Hi Steve,

    Thanks for the feedback on the WAS metrics app -- we spent a fair bit of time, with an early adopter from the CAB, making sure things would work as expected in z/OS environments. 

    On the request for native memory stats, are you looking for things equivalent to RSS / VSZ process stats?  The challenge for those will be providing them in a consistent way across all of our OS platforms and JVMs, so will take a bit of time to figure out if we can do it. 

    Seeing Claudia's note in this chain, I wanted to give a quick update.  While we haven't addressed every request, we've added the following capabilities to satisfy requests we'd heard during early demos and while working with early adopters.  These are in 8.5.5.20 and we expect will be in 9.0.5.9.  We hope to address more requests in subsequent releases.

    Improve metrics labelling

    -Add cell label to metrics

    Improve resiliency in cells with very slow servers

    -Add timeout for server metrics scraping


    Improve observability

    -Include server scrape duration metrics in metrics endpoint output

    Improve support for Kubernetes deployments

    -Provide property to enable/disable cell/node/server labels on WAS Prometheus output
    -Publish a Grafana dashboard with filtering options for namespaces / pods / containers


    ------------------------------
    Don Bourne
    WebSphere Observability Architect
    ------------------------------



  • 4.  RE: WAS metrics – Which enhancements would you like to see most?

    Posted Mon August 30, 2021 09:51 AM

    Don,

    Thanks for your quick response with regards to our request for z/OS Native Memory usage visibility.  Yes this would be similar to RSS / VSZ process stats.

    What we do today for Native Memory is periodically issue a display PID, parse the output dividing MAXMEMLIMIT PROCESS LIMIT into MAXMEMLIMIT CURRENT USAGE and save this as a percentage into a DB2 table for each server.

     

    Thanks

    Steve