Skip to content

Commit

Permalink
Revert "platform: Set RDMA protocol as default for trn1/trn1n platfor…
Browse files Browse the repository at this point in the history
…ms" (#603)

This reverts commit 756214e.

Commit 756214e was previously merged so that RDMA protocol became the
default protocol used on TRN1 instances. This commit was made as RDMA
compatible firmware was deployed on all TRN1 instances. However, in
order to be able to utilize the new RDMA compatible firmware, users are
required to initiate a reboot on their instances. This comes with the
side-effect that the plugin will fail to initialize on TRN1's that have
no been rebooted. So, this commit will set the default protocol to SEND
RECEIVE so that users can use this plugin out of the box without having
to reboot their instance.

Signed-off-by: Hunter North <hunnorth@amazon.com>
Signed-off-by: Nicholas Sielicki <nslick@amazon.com>
  • Loading branch information
hunnorth authored Sep 16, 2024
1 parent f1b91b7 commit b64107a
Showing 1 changed file with 2 additions and 2 deletions.
4 changes: 2 additions & 2 deletions src/platform-aws.c
Original file line number Diff line number Diff line change
Expand Up @@ -97,14 +97,14 @@ struct ec2_platform_data {
.name = "trn1.32xlarge",
.gdr_required = true,
.net_flush_required = true,
.default_protocol = "RDMA",
.default_protocol = "SENDRECV",
.domain_per_thread = 1,
},
{
.name = "trn1n.32xlarge",
.gdr_required = true,
.net_flush_required = true,
.default_protocol = "RDMA",
.default_protocol = "SENDRECV",
.domain_per_thread = 1,
}
};
Expand Down

0 comments on commit b64107a

Please sign in to comment.