risingwave playground adopts an in-memory SQLite meta store. After it runs for a while, the meta store database seems to be dropped, thus all data losses and the whole service fails.
It can happen at any random time, without any warning or the computer going to sleep beforehand.
According to SQLite:
The database is automatically deleted and memory is reclaimed when the last connection to the database closes.
I suppose we lose the connection to the in-memory SQLite database when the cluster is relatively idle. This might be due to the connection pool releasing resources according to certain rules.
2024-11-21T17:21:30.974444+08:00 INFO rw-standalone-meta bootstrap_recovery:failure_recovery{error=Hummock error: failed to access meta store: Query Error: error returned from database: (code: 1) no such table: hummock_sequence}: risingwave_meta::barrier::worker: recovery start!
2024-11-21T17:21:30.976676+08:00 WARN rw-standalone-compute risingwave_stream::task::barrier_manager: control stream reset with error error=gRPC request failed: Internal error: end of stream
2024-11-21T17:21:30.976746+08:00 WARN rw-standalone-compute risingwave_stream::task::barrier_manager: failed to notify reset of control stream
2024-11-21T17:21:31.196745+08:00 WARN rw-standalone-meta risingwave_meta::controller::cluster: Failed to load expire worker info from db error=Query Error: error returned from database: (code: 1) no such table: worker
2024-11-21T17:21:34.492494+08:00 ERROR rw-standalone-meta risingwave_meta::stream::source_manager: error happened while running source manager tick error=Query Error: error returned from database: (code: 1) no such table: fragment
2024-11-21T17:21:34.984923+08:00 WARN rw-standalone-meta risingwave_meta::hummock: Vacuum metadata error error=Hummock error: failed to access meta store: Execution Error: error returned from database: (code: 1) no such table: hummock_version_delta
2024-11-21T17:21:35.108002+08:00 WARN rw-standalone-meta risingwave_meta::hummock::manager::timer_task: failed to fetch created table ids error=Query Error: error returned from database: (code: 1) no such table: streaming_job
2024-11-21T17:21:35.275579+08:00 WARN rw-standalone-meta risingwave_meta::hummock::manager::compaction: Failed to get compaction task error=failed to access meta store: Query Error: error returned from database: (code: 1) no such table: table
risingwave playgroundadopts an in-memory SQLite meta store. After it runs for a while, the meta store database seems to be dropped, thus all data losses and the whole service fails.It can happen at any random time, without any warning or the computer going to sleep beforehand.
According to SQLite:
I suppose we lose the connection to the in-memory SQLite database when the cluster is relatively idle. This might be due to the connection pool releasing resources according to certain rules.