Github user srdo commented on a diff in the pull request:
https://github.com/apache/storm/pull/2761#discussion_r203693962
--- Diff:
storm-client/src/jvm/org/apache/storm/generated/ClusterSummary.java ---
@@ -63,8 +60,6 @@ public static _Fields findByThriftId(int fieldId) {
switch(fieldId) {
case 1: // SUPERVISORS
return SUPERVISORS;
- case 2: // NIMBUS_UPTIME_SECS
- return NIMBUS_UPTIME_SECS;
case 3: // TOPOLOGIES
return TOPOLOGIES;
--- End diff --
It's my impression that if we don't keep the field ids as they are now, it
won't be possible for people to do a rolling upgrade. E.g. if we changed
TOPOLOGIES to have id 2, then someone doing a rolling upgrade might have a
machine send ClusterSummary with TOPOLOGIES id 3 to an upgraded machine where
TOPOLOGIES has id 2, which breaks deserialization.
---
