Installation details
#ScyllaDB version: 5.2.13
#Cluster size:
os (RHEL/CentOS/Ubuntu/AWS AMI): CentOS
Is there any potential risk in large partitions or rows or cells if do compaction, or others?
Actually, a node is restarted with log “compaction_manager - compaction task … failed : std::bad_alloc(std::bad_alloc), will retry in 300 seconds”
Look for log lines for the same shard just before the bad_alloc. ScyllaDB will log when it’s starting a compaction task on a SSTable set, so you might have a pointer to which table/SSTables the problem is related to.
Also make sure you follow best practices in terms of vCPU/memory/storage to avoid bad_allocs: System Requirements | ScyllaDB Docs