[#MDEV-23647] Garbd can't initiate SST anymore in 10.5

The galera arbitrator binary garbd can be used to trigger any actual cluster node to act as a donor towards the arbitrator using the -~~sst and~~ -donor options.

Up to MariaDB 10.4 this works fine, but with 10.5 the SST request does not lead to the actual requested wsrep_sst_... script being executed anymore.

The donor node will just log:

Aug 21 15:10:27 node-1 mariadbd[2902]: 2020-08-21 15:10:27 0 [Note] WSREP: Member 0.0 (garbd) requested state transfer from 'node-1'. Selected 1.0 (node-1)(SYNCED) as donor.

Aug 21 15:10:27 node-1 mariadbd[2902]: 2020-08-21 15:10:27 0 [Note] WSREP: Shifting SYNCED -> DONOR/DESYNCED (TO: 3)

Aug 21 15:10:27 node-1 mariadbd[2902]: 2020-08-21 15:10:27 1 [Note] WSREP: ================================================

Aug 21 15:10:27 node-1 mariadbd[2902]: View:

Aug 21 15:10:27 node-1 mariadbd[2902]:   id: a0e49d83-e3bd-11ea-905f-ff2c5b67f9c2:3

Aug 21 15:10:27 node-1 mariadbd[2902]:   status: primary

Aug 21 15:10:27 node-1 mariadbd[2902]:   protocol_version: 4

Aug 21 15:10:27 node-1 mariadbd[2902]:   capabilities: MULTI-MASTER, CERTIFICATION, PARALLEL_APPLYING, REPLAY, ISOLATION, PAUSE, CAUSAL_READ, INCREMENTAL_WS, UNORDERED, PREORDERED, STREAMING, NBO

Aug 21 15:10:27 node-1 mariadbd[2902]:   final: no

Aug 21 15:10:27 node-1 mariadbd[2902]:   own_index: 1

Aug 21 15:10:27 node-1 mariadbd[2902]:   members(3):

Aug 21 15:10:27 node-1 mariadbd[2902]:         0: 71303bfe-e3c0-11ea-b367-d2112cd7bec7, garbd

Aug 21 15:10:27 node-1 mariadbd[2902]:         1: a0e2712c-e3bd-11ea-9f04-db21411dccbc, node-1

Aug 21 15:10:27 node-1 mariadbd[2902]:         2: d2324935-e3bd-11ea-a3b6-da0d74e66e47, node-2

Aug 21 15:10:27 node-1 mariadbd[2902]: =================================================

Aug 21 15:10:27 node-1 mariadbd[2902]: 2020-08-21 15:10:27 1 [Note] WSREP: wsrep_notify_cmd is not defined, skipping notification.

Aug 21 15:10:27 node-1 mariadbd[2902]: 2020-08-21 15:10:27 0 [Note] WSREP: 0.0 (garbd): State transfer from 1.0 (node-1) complete.

With 10.4 it shows the actual wsrep script being executed, but nothing like this is to be seen in the 10.5 donor log:

Aug 21 15:35:07 node-1 mysqld[2778]: 2020-08-21 15:35:07 0 [Note] WSREP: Running: 'wsrep_sst_backup --role 'donor' --address '' --socket '/var/run/mysqld/mysqld.sock' --datadir '/var/lib/mysql/' --gtid 'b21e7239-e3c1-11ea-a1cb-f36fe61c8c1c:0' --gtid-domain-id '0' --binlog 'node-1-bin' --mysqld-args --wsrep-new-cluster --wsrep_start_position=00000000-0000-0000-0000-000000000000:-1'

And on the garbd side:

2020-08-21 15:10:27.576  INFO: Shifting OPEN -> PRIMARY (TO: 3)

2020-08-21 15:10:27.577  INFO: Sending state transfer request: 'backup', size: 6

2020-08-21 15:10:27.578  INFO: Member 0.0 (garbd) requested state transfer from 'node-1'. Selected 1.0 (node-1)(SYNCED) as donor.

2020-08-21 15:10:27.578  INFO: Shifting PRIMARY -> JOINER (TO: 3)

2020-08-21 15:10:27.578  INFO: Closing send monitor...

2020-08-21 15:10:27.578  INFO: Closed send monitor.

2020-08-21 15:10:27.578  INFO: gcomm: terminating thread

2020-08-21 15:10:27.578  INFO: gcomm: joining thread

2020-08-21 15:10:27.579  INFO: gcomm: closing backend

2020-08-21 15:10:27.580  INFO: 0.0 (garbd): State transfer from 1.0 (node-1) complete.

2020-08-21 15:10:27.580  INFO: Shifting JOINER -> JOINED (TO: 3)

[MDEV-23647] Garbd can't initiate SST anymore in 10.5 Created: 2020-09-01 Updated: 2021-04-05 Resolved: 2021-01-21
Status:	Closed
Project:	MariaDB Server
Component/s:	Galera, Galera Arbitrator garbd, Galera SST
Affects Version/s:	10.5.5
Fix Version/s:	10.4.18, 10.5.9

[MDEV-23647] Garbd can't initiate SST anymore in 10.5 Created: 2020-09-01 Updated: 2021-04-05 Resolved: 2021-01-21