...
start coord
start master m1
start master m2
crash coord
start master m3
crash coord
start master m4
crash coord
start master m5
Works fine.
...
1.
start coord
start master m1
start master m2
crash master m2
...
Ryan's suggestion for a quick hack fix:
In BackupFailureMonitor.cc - in constructor for BackupFailureMonitor, change the initializer tracker(context, this) to tracker(context).
That worked for now.
...
2.
start coord: rc10
start master: rc01
start master: rc02
start master: rc03
crash master: rc02
...
<<< last two lines continue forever >>>
...
3.
start coord: rc10
start master: rc01
start master: rc02
start master: rc03
crash master: rc02
start master: rc04
...
rc04 gets entire server list (rc01, including rc02, rc03, rc04), in which rc02 has status crashed . why (status number: 1). Why isn't this server entirely gone?
...
4.
start coord: rc10
start master: rc01
start master: rc02
start master: rc03
crash master: rc02
start master: rc04 (as in point 3: gets entire server list, including rc07rc02, which has status crashed)
crash coord: rc10
restart coord: rc10
...