maoling commented on PR #2165: URL: https://github.com/apache/zookeeper/pull/2165#issuecomment-2176179886
Yes, I understand now. When you frequently create and delete numerous znodes at the first level, the write_per_namespace and read_per_namespace metrics will indeed expand and consume excessive memory. A possible workaround is to add an additional layer under the root directory of the tree (e.g., /taskName/taskId-[1,n]). While deleting the SummarySet key might work, it seems somewhat counterintuitive. Here are a few alternative suggestions (just my personal opinion): 1. Limit the size of a SummarySet. For example, when the number of keys exceeds 1000, prevent any further additions. 2. Implementing a blacklist is a good idea, as we have other metrics that are costly and can be resource-intensive to query at times. 2.1. Place the blacklist in the zoo.cfg file within the metrics section, instead of using a SystemProperty. 2.2. Ensure the blacklist supports all metric types, not just SummarySet. ``` ## Metrics Providers metricsProvider.exportJvmInfo=true # Support multiple and prefix matching metricsProvider.blackList=prep_processor*,sync_processor* ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
