[ https://issues.apache.org/jira/browse/KAFKA-13147?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Gary Russell updated KAFKA-13147: --------------------------------- Description: When the broker is down, the {code}connection-close-total{code} keeps incrementing with each connection attempt. This makes it more difficult to determine connection health. {code:java} System.out.println("Open count: " + metrics.entrySet().stream() .filter(entry -> entry.getKey().name().equals("connection-creation-total")) .map(entry -> entry.getValue().metricValue()) .collect(Collectors.toList())); System.out.println("Close count: " + metrics.entrySet().stream() .filter(entry -> entry.getKey().name().equals("connection-close-total")) .map(entry -> entry.getValue().metricValue()) .collect(Collectors.toList())); {code} {noformat} Open count: [1.0] Close count: [0.0] Open count: [3.0] Close count: [0.0] Open count: [3.0] Close count: [8.0] Open count: [3.0] Close count: [13.0] Open count: [3.0] Close count: [17.0] Open count: [3.0] Close count: [22.0] {noformat} It would also be more useful if metrics for connection counts by node were available. was: When the broker is down, the {code}connection-close-total{code} keeps incrementing with each connection event. This makes it more difficult to determine connection health. {code:java} System.out.println("Open count: " + metrics.entrySet().stream() .filter(entry -> entry.getKey().name().equals("connection-creation-total")) .map(entry -> entry.getValue().metricValue()) .collect(Collectors.toList())); System.out.println("Close count: " + metrics.entrySet().stream() .filter(entry -> entry.getKey().name().equals("connection-close-total")) .map(entry -> entry.getValue().metricValue()) .collect(Collectors.toList())); {code} {noformat} Open count: [1.0] Close count: [0.0] Open count: [3.0] Close count: [0.0] Open count: [3.0] Close count: [8.0] Open count: [3.0] Close count: [13.0] Open count: [3.0] Close count: [17.0] Open count: [3.0] Close count: [22.0] {noformat} It would also be more useful if metrics for connection counts by node were available. > connection-close-total Metric Keeps Incrementing While Broker Down > ------------------------------------------------------------------ > > Key: KAFKA-13147 > URL: https://issues.apache.org/jira/browse/KAFKA-13147 > Project: Kafka > Issue Type: Bug > Components: clients > Affects Versions: 2.8.0 > Reporter: Gary Russell > Priority: Major > > When the broker is down, the {code}connection-close-total{code} keeps > incrementing with each connection attempt. > This makes it more difficult to determine connection health. > {code:java} > System.out.println("Open count: " + metrics.entrySet().stream() > .filter(entry -> > entry.getKey().name().equals("connection-creation-total")) > .map(entry -> entry.getValue().metricValue()) > .collect(Collectors.toList())); > System.out.println("Close count: " + metrics.entrySet().stream() > .filter(entry -> entry.getKey().name().equals("connection-close-total")) > .map(entry -> entry.getValue().metricValue()) > .collect(Collectors.toList())); > {code} > {noformat} > Open count: [1.0] > Close count: [0.0] > Open count: [3.0] > Close count: [0.0] > Open count: [3.0] > Close count: [8.0] > Open count: [3.0] > Close count: [13.0] > Open count: [3.0] > Close count: [17.0] > Open count: [3.0] > Close count: [22.0] > {noformat} > It would also be more useful if metrics for connection counts by node were > available. -- This message was sent by Atlassian Jira (v8.3.4#803005)