disk_connections_hard_limit
Exception is thrown at a creation attempt when this limit is reached. Set to 0 to turn off hard limitation. The limit applies to the disks connections.disk_connections_rcvbuf
The size of the SO_RCVBUF option for disk (S3, Azure, GCS) connections. Defaults to 204800 (200 KB): object storage is reached intra-region where the bandwidth-delay product is well below 100 KB, so a small cap delivers full throughput while avoiding the multi-MiB per-socket buffers kernel autotuning would otherwise allocate. Set to 0 to use kernel TCP autotuning for the receive buffer instead. Note: changing this setting back to 0 restores autotuning only for newly created connections; existing pooled connections retain fixed buffer sizes until they are recreated.disk_connections_sndbuf
The size of the SO_SNDBUF option for disk (S3, Azure, GCS) connections. If set to a value greater than 0, overrides the kernel TCP autotuning for the send buffer. 0 = kernel default (autotuning). Note: changing this setting back to 0 restores autotuning only for newly created connections; existing pooled connections retain fixed buffer sizes until they are recreated.disk_connections_soft_limit
Connections above this limit have significantly shorter time to live. While the number of connections in the group, in use and idle together, is at or above this limit, reads from S3 disks stop keeping a connection open between buffer fills: each fill is a separate request whose connection returns to the pool as soon as the fill completes. This bounds the connections held by readers that keep many streams open at once, such as merges of parts with aJSON column, at the cost of one request per fill. The limit applies to the disks connections.