]
Radim Vansa commented on ISPN-4949:
-----------------------------------
[~belaban] You're right that it limits scalability - I don't say this has to be
the default, but it is still a responsibility of JGroups (transport layer) to manage group
membership.
I think that we can't workaround the need for some consensus on 'who's the
coordinator' with current Infinispan architecture - at best we could somehow limit the
need for updates 'who's in the view' and keep that info only in coordinator.
Still, I think that it could work well even with limited scalability. In production there
would be tens of thousands of messages per second per node, therefore, few one
shouldn't matter. And when the cluster goes wild with network, degraded performance is
to be expected.
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