Description of problem: Dist-geo-rep : 'gluster volume geo <master_vol> <slave_ip>::<slave_vol> delete' command deletes all session having that slave_vol name ; even though those session are created using different slave_ip from the same salve cluster Version-Release number of selected component (if applicable): always How reproducible: 3.4.0.13rhs-1.el6rhs.x86_64 Steps to Reproduce: 1. slave cluster has 4 servers. rhsauto018 rhsauto031 rhsauto026 rhsauto027 2. create session using rhsauto031 gluster volume geo test5 rhsauto031.lab.eng.blr.redhat.com::slave5 status NODE MASTER SLAVE HEALTH UPTIME ----------------------------------------------------------------------------------------------------------------- DVM4.lab.eng.blr.redhat.com test5 rhsauto031.lab.eng.blr.redhat.com::slave5 Not Started N/A DVM3.lab.eng.blr.redhat.com test5 rhsauto031.lab.eng.blr.redhat.com::slave5 Not Started N/A DVM5.lab.eng.blr.redhat.com test5 rhsauto031.lab.eng.blr.redhat.com::slave5 Not Started N/A DVM6.lab.eng.blr.redhat.com test5 rhsauto031.lab.eng.blr.redhat.com::slave5 Not Started N/A DVM1.lab.eng.blr.redhat.com test5 rhsauto031.lab.eng.blr.redhat.com::slave5 Not Started N/A DVM2.lab.eng.blr.redhat.com test5 rhsauto031.lab.eng.blr.redhat.com::slave5 Not Started N/A 3. now create session using another IP from slave cluster but volume name should be same and start that session [root@DVM4 geo-replication]# gluster volume geo test5 rhsauto018.lab.eng.blr.redhat.com::slave5 status NODE MASTER SLAVE HEALTH UPTIME ------------------------------------------------------------------------------------------------------------ DVM4.lab.eng.blr.redhat.com test5 rhsauto018.lab.eng.blr.redhat.com::slave5 faulty N/A DVM2.lab.eng.blr.redhat.com test5 rhsauto018.lab.eng.blr.redhat.com::slave5 faulty N/A DVM3.lab.eng.blr.redhat.com test5 rhsauto018.lab.eng.blr.redhat.com::slave5 faulty N/A DVM6.lab.eng.blr.redhat.com test5 rhsauto018.lab.eng.blr.redhat.com::slave5 faulty N/A DVM5.lab.eng.blr.redhat.com test5 rhsauto018.lab.eng.blr.redhat.com::slave5 faulty N/A DVM1.lab.eng.blr.redhat.com test5 rhsauto018.lab.eng.blr.redhat.com::slave5 faulty N/A 4. delete the session created in step 2 and verify that it has also deleted session created in step 3 [root@DVM4 geo-replication]# gluster volume geo test5 rhsauto031.lab.eng.blr.redhat.com::slave5 delete Deleting geo-replication session between test5 & rhsauto031.lab.eng.blr.redhat.com::slave5 has been successful [root@DVM4 geo-replication]# gluster volume geo test5 status No active geo-replication sessions for test5 Actual results: deleting all sessions having that slave volume name Expected results: should not delete all session Additional info:
The geo-replication in the new code works with cluster-id, and not exactly on the 'IP/hostname' of the slave node. 'IP/hostname' of the slave is required just to fetch the information of the remote cluster, and afterwords the housekeeping all happens on cluster-id (an uuid). If the IP address is changed, but the slave cluster is same, as per geo-replication, its still the same session, and hence it deletes both the sessions.