[ 
https://issues.apache.org/jira/browse/KNOX-2899?focusedWorklogId=857647&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-857647
 ]

ASF GitHub Bot logged work on KNOX-2899:
----------------------------------------

                Author: ASF GitHub Bot
            Created on: 18/Apr/23 11:23
            Start Date: 18/Apr/23 11:23
    Worklog Time Spent: 10m 
      Work Description: smolnar82 commented on PR #749:
URL: https://github.com/apache/knox/pull/749#issuecomment-1512906616

   For reviewers: I wanted to do the least intrusive change possible, that's 
why the private method and the `includedServices` parameter are still in place 
and in use. Since everything happens in memory, it's not a real performance 
issue.




Issue Time Tracking
-------------------

    Worklog Id:     (was: 857647)
    Time Spent: 20m  (was: 10m)

> Disable service-based discovery filter
> --------------------------------------
>
>                 Key: KNOX-2899
>                 URL: https://issues.apache.org/jira/browse/KNOX-2899
>             Project: Apache Knox
>          Issue Type: Task
>          Components: Server
>            Reporter: Sandor Molnar
>            Assignee: Sandor Molnar
>            Priority: Major
>             Fix For: 2.1.0
>
>          Time Spent: 20m
>  Remaining Estimate: 0h
>
> With KNOX-2680, the Knox team introduced two major changes in Cloudera 
> Manager service discovery:
>  * the first is that Knox only discovers services that are actually listed in 
> the given descriptor (instead of discovering the entire cluster)
>  * from now on, Knox maintains an in-memory repository of discovered data 
> (cluster, service, and role configs). Caffeine is used as the cache 
> implementation with the default of 10 mins of cache entry TTL.
> It turns out the first improvement may cause redundant topology redeployments 
> due to "missing" service discovery details while the {{Polling Configuration 
> Analyzer}} finds a relevant event that belongs to a service that was not 
> included in the latest discovery round.
> For now, we should disable that feature temporarily until it's clear how to 
> enable and address the redundant redeployment issue.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to