The problem still exists with 2.10.6 [2 controller scenario], obvious it
seems only to occur if I do a forced crash of a controller via command
"killall java":
Controller A - 192.168.10.94
Controller B - 192.168.10.90
If I do a forced crash of "controller A" via "killall java" and then
restart this "controller A", he is rejoining the group. In the logs I
can see, that Controller B had realized that Controller A has failed.
But a show controller lists furthermore both controllers. If I now
restart Controller A, he is rejoining the group, but "controller B"
output shows now 2 times controller A as group member:
Ready to administrate virtual database qualiproof-dev
qualiproof-dev(admin) > show controllers
qualiproof-dev is hosted by 3 controller(s):
192.168.10.94:1090 <-- "old" died controller 192.168.10.94
192.168.10.94:1090 <-- "new" living controller 192.168.10.94
192.168.10.90:1090
In the log I see following group-membership:
2007-03-01 12:52:35,542 INFO controller.virtualdatabase.qualiproof-dev
Member(address=/192.168.10.90:60421, uid=192.168.10.90:60421) see
members:[Member(address=/192.168.10.90:60421, uid=192.168.10.90:60421),
Member(address=/192.168.10.94:32776, uid=192.168.10.94:32776)] and has
mapping:{Member(address=/192.168.10.94:36233,
uid=192.168.10.94:36233)=192.168.10.94:1090,
Member(address=/192.168.10.94:32776,
uid=192.168.10.94:32776)=192.168.10.94:1090,
Member(address=/192.168.10.90:60421,
uid=192.168.10.90:60421)=192.168.10.90:1090}
Especially this
has mapping:{Member(address=/192.168.10.94:36233,
uid=192.168.10.94:36233)=192.168.10.94:1090,
Member(address=/192.168.10.94:32776,
uid=192.168.10.94:32776)=192.168.10.94:1090,
Member(address=/192.168.10.90:60421,
uid=192.168.10.90:60421)=192.168.10.90:1090}
seems for me not to be correct? The ip 192.168.10.94 is only ONE
Controller and should only be once listed. Furthermore a transfer dump
to this controller is not possible anymore - It simply hangs endless.
Remember: this problem only occurs, if I crash the controller via
"killall java" (simulating crash of the controller process). If I simply
power off the server, there is no problem.
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
