nodetool reports the uptime of scylla-jmx and not the uptime of scylla server

[tarzanek]

If you look here:
https://github.com/scylladb/scylla-tools-java/blob/master/src/java/org/apache/cassandra/tools/nodetool/Info.java#L59

you will see we get "Uptime (seconds)" from a jmx mxbean of java runtime for service scylla-jmx, not the one from scylla-server
so nodetool reports misleading/wrong info.
(scylla-jmx service can be restarted and the time will be wrong)

[tarzanek]

fwiw
same thing is for "Heap Memory (MB)", it also comes from jmx

[tarzanek]

https://github.com/scylladb/scylla/blob/master/api/api-doc/system.json#L56
REST system /system/uptime_ms now provides uptime

[avikivity]

haha.

[tarzanek]

https://github.com/scylladb/scylla-tools-java/blob/master/src/java/org/apache/cassandra/tools/NodeProbe.java#L632
is the jmx call, unfortunately we don't have any runtimeProxy override in our scylla-jmx repo so this asks for a Scylla REST NodeProbe as in #100

[tarzanek]

this was partially (fully?) fixed by
scylladb/scylla-jmx#155

[tarzanek]

@amnonh can this issue be closed?

[tarzanek]

hmm, actually "Heap Memory (MB)" is likely not corresponding to scylla mem usage