f6ac969f1e
Before stopping the db itself, stop the migration service. It must be stopped before RPC, but RPC is not stopped yet itself, so we should be safe here. Here's the tail of the resulting logs: INFO 2019-11-20 11:22:35,193 [shard 0] init - shutdown migration manager INFO 2019-11-20 11:22:35,193 [shard 0] migration_manager - stopping migration service INFO 2019-11-20 11:22:35,193 [shard 1] migration_manager - stopping migration service INFO 2019-11-20 11:22:35,193 [shard 0] init - Shutdown database started INFO 2019-11-20 11:22:35,193 [shard 0] init - Shutdown database finished INFO 2019-11-20 11:22:35,193 [shard 0] init - stopping prometheus API server INFO 2019-11-20 11:22:35,193 [shard 0] init - Scylla version 666.development-0.20191120.25820980f shutdown complete. Also -- stop the mm on drain before the commitlog it stopped. [Tomasz: mm needs the cl because pulling schema changes from other nodes involves applying them into the database. So cl/db needs to be stopped after mm is stopped.] The drain logs would look like ... INFO 2019-11-25 11:00:40,562 [shard 0] migration_manager - stopping migration service INFO 2019-11-25 11:00:40,562 [shard 1] migration_manager - stopping migration service INFO 2019-11-25 11:00:40,563 [shard 0] storage_service - DRAINED: and then on stop ... INFO 2019-11-25 11:00:46,427 [shard 0] init - shutdown migration manager INFO 2019-11-25 11:00:46,427 [shard 0] init - Shutdown database started INFO 2019-11-25 11:00:46,427 [shard 0] init - Shutdown database finished INFO 2019-11-25 11:00:46,427 [shard 0] init - stopping prometheus API server INFO 2019-11-25 11:00:46,427 [shard 0] init - Scylla version 666.development-0.20191125.3eab6cd54 shutdown complete. Fixes #5300 Signed-off-by: Pavel Emelyanov <xemul@scylladb.com> Message-Id: <20191125080605.7661-1-xemul@scylladb.com>