]
Bela Ban commented on ISPN-4949:
--------------------------------
Re consensus based view installation: JGroups doesn't use consensus for view
installation, because it's simply faster to do without consensus, which includes round
trips at a latency and a timeout. For large clusters, this won't be scalable. Imagine
you need to install a new view in 500 nodes. This would include invoking ~ 500 RPCs, wait
for all responses (or a timeout) and then do a second RPC committing the proposed view.
Additional logic would be needed to handle dangling prepare and commit, ie. when the
leader crashes after the PREPARE or COMMIT phase, plus possible vote collection when a new
coord takes over.
I'll look into this though, so I created [1] as a result.
[1] JGRP-1901
Split brain: inconsistent data after merge
------------------------------------------
Key: ISPN-4949
URL:
https://issues.jboss.org/browse/ISPN-4949
Project: Infinispan
Issue Type: Bug
Components: State Transfer
Affects Versions: 7.0.0.Final
Reporter: Radim Vansa
Priority: Critical
1) cluster A, B, C, D splits into 2 parts:
A, B (coord A) finds this out immediately and enters degraded mode with CH [A, B, C, D]
C, D (coord D) first detects that B is lost, gets view A, C, D and starts rebalance with
CH [A, C, D]. Segment X is primary owned by C (it had backup on B but this got lost)
2) D detects that A was lost as well, therefore enters degraded mode with CH [A, C, D]
3) C inserts entry into X: all owners (only C) is present, therefore the modification is
allowed
4) cluster is merged and coordinator finds out that the max stable topology has CH [A, B,
C, D] (it is the older of the two partitions' topologies, got from A, B) - logs
'No active or unavailable partitions, so all the partitions must be in degraded
mode' (yes, all partitions are in degraded mode, but write has happened in the
meantime)
5) The old CH is broadcast in newest topology, no rebalance happens
6) Inconsistency: read in X may miss the update