From d64af465d73ea7478bed7a8458ebf9ffab402bb8 Mon Sep 17 00:00:00 2001 From: Tariq Ibrahim Date: Thu, 15 Feb 2024 15:46:37 -0800 Subject: [PATCH] wait for mofed driver installation after the cleanup driver call Signed-off-by: Tariq Ibrahim --- driver-manager | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/driver-manager b/driver-manager index 95d613dd..464928a5 100755 --- a/driver-manager +++ b/driver-manager @@ -571,12 +571,6 @@ _wait_for_mofed_driver() { } uninstall_driver() { - # when GPUDirectRDMA is enabled, wait until MOFED driver has finished installing - if _gpu_direct_rdma_enabled; then - echo "GPUDirectRDMA is enabled, validating MOFED driver installation" - _wait_for_mofed_driver - fi - # don't attempt to un-install if driver is pre-installed on the node if _host_driver; then echo "NVIDIA GPU driver is already pre-installed on the node, disabling the containerized driver on the node" @@ -645,6 +639,12 @@ uninstall_driver() { _exit_failed fi + # when GPUDirectRDMA is enabled, wait until MOFED driver has finished installing + if _gpu_direct_rdma_enabled; then + echo "GPUDirectRDMA is enabled, validating MOFED driver installation" + _wait_for_mofed_driver + fi + if _is_gpu_pod_eviction_enabled || _is_auto_drain_enabled; then # uncordon the node in case if the pod has restarted abruptly after we cordoned the node _uncordon_k8s_node