[OAK-3419] ClusterNodeInfo.createInstance fails to clean up random entries

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.0.19, 1.2.4, 1.3.5
• Fix Version/s: 1.0.22, 1.2.7, 1.3.7, 1.4
• Component/s: core
• Labels: resilience
• Assignee: Julian Reschke
• Reporter: Julian Reschke
• Created: 17/Sep/15 16:17
• Updated: 02/Mar/16 14:23
• Resolved: 18/Sep/15 15:21

Description

            String mId = "" + doc.get(MACHINE_ID_KEY);
            String iId = "" + doc.get(INSTANCE_ID_KEY);
            if (machineId.startsWith(RANDOM_PREFIX)) {
                // remove expired entries with random keys
                store.remove(Collection.CLUSTER_NODES, key);
                continue;
            }

The intent seems to be to cleanup entries in the cluster node table that start with RANDOM_PREFIX. However, machineId is checked instead of mId. When createInstance is called with a random id, the whole table might get wiped out.