tmp: dont use async engine

Signed-off-by: Patrick Roy <[email protected]>

Author: roypat

tests/integration_tests/performance/test_block_ab.py

@@ -60,7 +60,6 @@ def run_fio(microvm, mode, block_size):
         .with_arg("--randrepeat=0")
         .with_arg(f"--bs={block_size}")
         .with_arg(f"--size={BLOCK_DEVICE_SIZE_MB}M")
-        .with_arg("--ioengine=libaio")
         .with_arg("--iodepth=32")
         # Set affinity of the entire fio process to a set of vCPUs equal in size to number of workers
         .with_arg(

tools/ab_test.py

@@ -46,13 +46,6 @@
     {"instance": "m6a.metal", "performance_test": "test_network_tcp_throughput"},
     # Network throughput on m7a.metal
     {"instance": "m7a.metal-48xl", "performance_test": "test_network_tcp_throughput"},
-    # block device latency metrics on ARM
-    {"instance": "m6g.metal", "metric": "clat_read"},
-    {"instance": "m6g.metal", "metric": "clat_write"},
-    {"instance": "m7g.metal", "metric": "clat_read"},
-    {"instance": "m7g.metal", "metric": "clat_write"},
-    {"instance": "m8g-24xl.metal", "metric": "clat_read"},
-    {"instance": "m8g-48xl.metal", "metric": "clat_write"},
 ]