Originally from the User Slack
@AmiSri: I am using Scylla version 5.4.3
I have 5882 tables till now
And recently I faced std::bad_alloc and writes failed
Is there a limit of number of tables in a keyspace
Or limit of total number of tables in a cluster
How to overcome the issue??
@Piotr_Smaroń: It may be that you created too many tables within one KS, per the doc: https://opensource.docs.scylladb.com/master/reference/limits.html, we “support” up to 5k tablets per KS
Limits | ScyllaDB Docs
@AmiSri: Yes its in a single keyspace with cdc enabled
Is this relevant for all versions of ScyllaDB??
@Piotr_Smaroń
@Piotr_Smaroń: we’re not testing limits in each version, but you can assume they apply to the recent versions
@AmiSri: On dropping the tables, i found load of only one node is seems to be reduced while load on all other nodes in the cluster is same as earlier.
I have cleared snapshots too
Is a repair or rebuild required??
What is the correct way to drop tables in a cluster??
@Piotr_Smaroń
@Piotr_Smaroń: just DROP TABLE should be enough
if the disk is still full, you may take a look here https://opensource.docs.scylladb.com/stable/troubleshooting/drop-table-space-up.html
Dropping tables doesn’t mean your load will decrease, for load to decrease you must lower the number of queries to the db. Perhaps the tables you dropped had partitions located only on a single node, and since you stopped querying these, your load for this single node has dropped
Dropped (or truncated) Table (or keyspace) and Disk Space is not Reclaimed | ScyllaDB Docs
@AmiSri: This is a restore server and no quering is going on here, do you mean partition key
Please help me understanding “number of queries” mean
@Piotr_Smaroń: by queries I meant both reading and writing to a database