[ https://issues.apache.org/jira/browse/CASSANDRA-8252?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14314492#comment-14314492 ]
Shawn Kumar commented on CASSANDRA-8252: ---------------------------------------- While trying to modify replace_address_test.py to check system.peers, I had some difficulty due to unexpected results. I committed a modified version of replace_address_test with some debug statements [here|https://github.com/riptano/cassandra-dtest/blob/topology/replace_address_test.py] to illustrate the changes in the table through the process. Issues I am having are: variable system.peers table after replacing across runs (although the replaced address does not appear, it seems rare that the replacing node is noticed by 1 and 2), upon trying to truncate the table and restart nodes I can occasionally see the replaced address (in this case 127.0.0.3). > dtests that involve topology changes should verify system.peers on all nodes > ---------------------------------------------------------------------------- > > Key: CASSANDRA-8252 > URL: https://issues.apache.org/jira/browse/CASSANDRA-8252 > Project: Cassandra > Issue Type: Test > Components: Tests > Reporter: Brandon Williams > Assignee: Shawn Kumar > Fix For: 2.1.3, 2.0.13 > > > This is especially true for replace where I've discovered it's wrong in > 1.2.19, which is sad because now it's too late to fix. We've had a lot of > problems with incorrect/null system.peers, so after any topology change we > should verify it on every live node when everything is finished. -- This message was sent by Atlassian JIRA (v6.3.4#6332)