The test items are: https://github.com/openucx/sparkucx
Test environment: We set up a Hadoop cluster with 2 machines. The RDMA network card model is Mellanox Technologies MT27800 Family [ConnectX-5]
Hadoop version: 3.2.1
Spark version: 3.0.0
When we applied the sparkucx plugin, we tested the wordcount sample in HiBench and found the following error printed in the Spark task log:
proto_select.c:232 Assertion init_params.rkey_config_key->ep_cfg_index == ep_cfg_index failed: rkey->ep_cfg_index=0 ep_cfg_index=4
The specific error stack and RDMA network card information are attached
The test items are: https://github.com/openucx/sparkucx Test environment: We set up a Hadoop cluster with 2 machines. The RDMA network card model is Mellanox Technologies MT27800 Family [ConnectX-5] Hadoop version: 3.2.1 Spark version: 3.0.0
When we applied the sparkucx plugin, we tested the wordcount sample in HiBench and found the following error printed in the Spark task log: proto_select.c:232 Assertion
init_params.rkey_config_key->ep_cfg_index == ep_cfg_index
failed: rkey->ep_cfg_index=0 ep_cfg_index=4The specific error stack and RDMA network card information are attached