You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
yb-master UI at 7000 showed node3 is the leader. Then did a kill -9 on the this master process. The logs showed that master-1 became the new leader, but that master-1 process crashed when its UI was accessed.
Current theory is that GetMasterEntryForHost returned error status semantics regressed recently and are causing this codepath in master-path-handlers.cc not being taken on failures.
…rHosts
Summary:
We ignore controller status in GetMasterEntryForHosts, in this case response is empty and its http_addresses is also empty.
Added handling of controller status to address this issue.
Test Plan:
Add 127.0.0.1 as node1, and same for 2 & 3, in /etc/hosts.
Start masters of a local RF=3 cluster (on Mac) using:
```
./yb-master --webserver_interface 127.0.0.1 --rpc_bind_addresses=127.0.0.1 --server_broadcast_addresses=node1:7100 --use_private_ip=zone --master_addresses node1:7100,node2:7100,node3:7100 --fs_data_dirs "/tmp/yblocal1/" >& /tmp/yb-master_1.out &
./yb-master --webserver_interface 127.0.0.2 --rpc_bind_addresses=127.0.0.2 --server_broadcast_addresses=node2:7100 --use_private_ip=zone --master_addresses node1:7100,node2:7100,node3:7100 --fs_data_dirs "/tmp/yblocal2/" >& /tmp/yb-master_2.out &
./yb-master --webserver_interface 127.0.0.3 --rpc_bind_addresses=127.0.0.3 --server_broadcast_addresses=node3:7100 --use_private_ip=zone --master_addresses node1:7100,node2:7100,node3:7100 --fs_data_dirs "/tmp/yblocal3/" >& /tmp/yb-master_3.out &
```
yb-master UI at 7000 shows the leader. kill -9 this master process.
New master leader should not crash when its UI is accessed.
Reviewers: robert, mikhail, bharat
Reviewed By: bharat
Subscribers: ybase, bogdan
Differential Revision: https://phabricator.dev.yugabyte.com/D5441
Added 127.0.01 as node1, and same for 2 & 3, in /etc/hosts.
Then started masters of a local RF=3 cluster (on Mac) using:
yb-master UI at 7000 showed node3 is the leader. Then did a
kill -9
on the this master process. The logs showed that master-1 became the new leader, but that master-1 process crashed when its UI was accessed.The text was updated successfully, but these errors were encountered: