nixl backend issue for kv offloading

[navmarri14]

Hi, I am seeing the following warning when using nixl backend for kv cache offloading. I double checked that the nixl is installed correctly. am I missing something?

createXferReq: no specified or potential backend had the required registrations to be able to do the transfer

sg lang launch command

python3 -m sglang.launch_server \
  --model-path /mnt/models \
  --tp-size 4 \
  --tool-call-parser glm45 \
  --reasoning-parser glm45 \
  --speculative-algorithm NEXTN \
  --speculative-num-steps 3 \
  --speculative-eagle-topk 1 \
  --speculative-num-draft-tokens 4 \
  --page-size 128 \
  --enable-flashinfer-allreduce-fusion \
  --enable-fused-qk-norm-rope \
  --host 0.0.0.0 \
  --port 80 \
  --enable-hierarchical-cache \
  --hicache-ratio 2
  --hicache-write-policy write_through
  --hicache-mem-layout page_first_direct
  --hicache-storage-backend nixl