Description of problem: ======================= On a default volume there is a volume option 'transport.address-family: inet' which vanishes on resetting a volume and because of this change in volume options ,brick mux feature breaks Version-Release number of selected component (if applicable): ============================================================ Have been seeing this from a long time Current testing build - glusterfs-server-6.0-24.el7rhgs.x86_64 How reproducible: ================= Everytime Steps to Reproduce: 1.Enable brickmux and create 2-3 volumes , start the volume , all will have same PID and volume option transport.address-family: inet 2.Reset a volume, this volume option is no longer present 3.Start - Stop the volume which was reset Actual results: =============== The volume which was reset and start-stopped , comes up with a different PID Expected results: ================= If on a new volume option transport.address-family: inet is present it should not disappear after volume reset Additional info: ================= [root@rhs-gp-srv1 glusterd]# gluster v info Volume Name: disperse-vol Type: Distributed-Disperse Volume ID: e4741d05-6bb0-4389-8e24-1d556cc01ae8 Status: Started Snapshot Count: 0 Number of Bricks: 2 x (4 + 2) = 12 Transport-type: tcp Bricks: Brick1: 10.70.36.8:/gluster/brick1/ec1 Brick2: 10.70.36.5:/gluster/brick1/ec1 Brick3: 10.70.36.6:/gluster/brick1/ec1 Brick4: 10.70.36.9:/gluster/brick1/ec1 Brick5: 10.70.36.17:/gluster/brick1/ec1 Brick6: 10.70.36.15:/gluster/brick1/ec1 Brick7: 10.70.36.8:/gluster/brick2/ec2 Brick8: 10.70.36.5:/gluster/brick2/ec2 Brick9: 10.70.36.6:/gluster/brick2/ec2 Brick10: 10.70.36.9:/gluster/brick2/ec2 Brick11: 10.70.36.17:/gluster/brick2/ec2 Brick12: 10.70.36.15:/gluster/brick2/ec2 Options Reconfigured: storage.fips-mode-rchecksum: on nfs.disable: on cluster.brick-multiplex: enable Volume Name: ec-vol Type: Disperse Volume ID: a6c976c6-a705-4782-8684-c2e60b805b78 Status: Started Snapshot Count: 0 Number of Bricks: 1 x (4 + 2) = 6 Transport-type: tcp Bricks: Brick1: 10.70.36.8:/gluster/brick3/ec1 Brick2: 10.70.36.5:/gluster/brick3/ec1 Brick3: 10.70.36.6:/gluster/brick3/ec1 Brick4: 10.70.36.9:/gluster/brick3/ec1 Brick5: 10.70.36.17:/gluster/brick3/ec1 Brick6: 10.70.36.15:/gluster/brick3/ec1 Options Reconfigured: transport.address-family: inet storage.fips-mode-rchecksum: on nfs.disable: on cluster.brick-multiplex: enable Volume Name: ec-vol1 Type: Disperse Volume ID: 6c7f409c-6fca-4d08-8aac-da7fb06d6ec4 Status: Started Snapshot Count: 0 Number of Bricks: 1 x (4 + 2) = 6 Transport-type: tcp Bricks: Brick1: 10.70.36.8:/gluster/brick4/ec1 Brick2: 10.70.36.5:/gluster/brick4/ec1 Brick3: 10.70.36.6:/gluster/brick4/ec1 Brick4: 10.70.36.9:/gluster/brick4/ec1 Brick5: 10.70.36.17:/gluster/brick4/ec1 Brick6: 10.70.36.15:/gluster/brick4/ec1 Options Reconfigured: transport.address-family: inet storage.fips-mode-rchecksum: on nfs.disable: on cluster.brick-multiplex: enable Volume disperse-vol has been reset [root@rhs-gp-srv1 glusterd]# gluster v get disperse-vol all|grep 'transport' transport.keepalive 1 transport.listen-backlog 1024 [root@rhs-gp-srv1 glusterd]# [root@rhs-gp-srv1 glusterd]# gluster v get ec-vol1 all |grep transport transport.keepalive 1 transport.listen-backlog 1024 transport.address-family inet
Here, the brick-mux feature failing because of the option change. When different options are set for volumes, the volumes will not be compatible with each other and we will have different PID's for bricks. The actual issue here is: the option <transport.address-family> should not get disappear after volume reset. I will look into it.
upstream patch: https://review.gluster.org/#/c/glusterfs/+/23921/
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory (glusterfs bug fix and enhancement update), and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2020:5603