-
1. Re: Incorrect cluster re-balance by reconnecting node after network outage
rvansa Jul 1, 2016 8:06 AM (in response to sjaiswal)Do you have partition handling enabled? With this setting, the node should wipe out its data when re-joining the cluster (when it sees that it has been removed from the cluster as inactive).
-
2. Re: Incorrect cluster re-balance by reconnecting node after network outage
sjaiswal Jul 1, 2016 11:26 AM (in response to rvansa)I have tried that too Radim.
The behavior was after rejoining of main cluster back it has also wiped data on original members.
this was checked without versioning, will update with behavior again with versioning enabled as simple and partition handling enabled.
Thanks for a lead.
-
3. Re: Incorrect cluster re-balance by reconnecting node after network outage
rvansa Jul 7, 2016 4:54 AM (in response to sjaiswal)Data are wiped out only on the non-available partition when joining to available partition. If you see different behaviour, it's a bug. Could you write a reproducer? infinispan/PartitionHappeningTest.java at 7fa26cef1c9189c36da7694fb3f0ff0acf3bfd82 · infinispan/infinispan · GitHub should give you some clues how to simulate a partition.
-
4. Re: Incorrect cluster re-balance by reconnecting node after network outage
wdfink Jul 7, 2016 2:15 PM (in response to sjaiswal)With PartitionHandling enabled it should not wipe data from an AVAILABLE partition. If there is no partition like this (all are DEGRADED) none of the partitions will be wiped-out, it will merge all data.
This is according to the PartitionHandling definition.
If there is something unexpected please show the case and provide logfiles