_read_concurrency_sem: restricted mutation reader queue

[Jean Carlo]

Hello folks,

I am facing some cluster overload for a pretty low throughput in a cluster scylla of 24 nodes and 3 racks.

I am having a table with an index and I am doing some test including updates in the base table that fire an update in the index.

I can see in my scylla logs a messages:

[shard 12] storage_proxy - Exception when communicating with 10.7.214.228, to read from user.keychain_keyid_index: std::runtime_error (_read_concurrency_sem: restricted mutation reader queue overload)

 [shard 15] storage_proxy - Exception when communicating with 10.7.214.228, to read from user.keychain: std::runtime_error (_read_concurrency_sem: restricted mutation reader queue overload)

So although I am not using a MV, it looks pretty similar to the issue

MV updates are issued with an infinite timeout - a recepy for a live-lock #10031

Which is still open. I wanted to confirm with you guys that I am effectively hitting this issue

Thx