Skip to content

Commit

Permalink
frontera: config options for tcp and sockets (#127)
Browse files Browse the repository at this point in the history
- parse provider in get_results.py
- add placeholder tcp and sockets env vars

Signed-off-by: Dalton Bohning <dalton.bohning@intel.com>
  • Loading branch information
daltonbohning authored Apr 22, 2022
1 parent d45f949 commit 810d6a7
Show file tree
Hide file tree
Showing 3 changed files with 50 additions and 5 deletions.
5 changes: 5 additions & 0 deletions frontera/daos_server.yml
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,8 @@ name: daos_server
nr_hugepages: 4096
port: 10001
provider: ofi+verbs;ofi_rxm
#provider: ofi+tcp;ofi_rxm
#provider: ofi+sockets
crt_timeout: 180
crt_ctx_share_addr: 0
engines:
Expand All @@ -25,6 +27,9 @@ engines:
- SWIM_PROTOCOL_PERIOD_LEN=2000
- SWIM_SUSPECT_TIMEOUT=19000
- SWIM_PING_TIMEOUT=1900
# For sockets
#- FI_SOCKETS_MAX_CONN_RETRY=1
#- FI_SOCKETS_CONN_TIMEOUT=2000
fabric_iface: ib0
fabric_iface_port: 31416
first_core: 0
Expand Down
16 changes: 13 additions & 3 deletions frontera/env_daos
Original file line number Diff line number Diff line change
Expand Up @@ -3,11 +3,21 @@ export FI_OFI_RXM_USE_SRX=1
export FI_MR_CACHE_MONITOR=disabled
export D_LOG_FILE=/tmp/daos_logs/daos_client.log
export D_LOG_MASK=ERR
export CRT_PHY_ADDR_STR="ofi+verbs;ofi_rxm"
export OFI_INTERFACE=ib0
export OFI_DOMAIN=mlx5_0
export DAOS_AGENT_DRPC_DIR=/tmp/daos_agent
export DAOS_DISABLE_REQ_FWD=1
export OFI_INTERFACE=ib0

# For verbs
export CRT_PHY_ADDR_STR="ofi+verbs;ofi_rxm"
export OFI_DOMAIN=mlx5_0

# For tcp
#export CRT_PHY_ADDR_STR="ofi+tcp;ofi_rxm"
#export OFI_DOMAIN=ib0

# For sockets
#export CRT_PHY_ADDR_STR="ofi+sockets"
#export OFI_DOMAIN=ib0

LOCATION=$1

Expand Down
34 changes: 32 additions & 2 deletions frontera/get_results.py
Original file line number Diff line number Diff line change
Expand Up @@ -187,6 +187,32 @@ def get_num_targets(output_file_path, slurm_job_id):
return None
return match.group(1)

def get_provider(output_file_path, slurm_job_id):
"""Get the provider from the server config.
Args:
output_file_path (str): path to the log output.
slurm_job_id (str): the slurm job id.
Returns:
str: the number of targets
None on failure.
"""
if not slurm_job_id:
return None

dir_name = dirname(output_file_path)

config_path = join(dir_name, "daos_server.yml")
config = read_file(config_path)
if not config:
return None
match = re.search("^provider: (.*)$", config, re.MULTILINE)
if not match:
return None
return match.group(1)

def get_mdtest_metric_max(metric, output):
"""Get the "max" for an mdtest metric.
Expand Down Expand Up @@ -500,6 +526,7 @@ def __init__(self, csv_file_path, output_style="full"):
"start_time": "Date",
"end_time": "End",
"daos_commit": "Commit",
"provider": "Provider",
"oclass": "Oclass",
"num_servers": "Num_Servers",
"num_targets": "Num Targets",
Expand All @@ -518,7 +545,7 @@ def __init__(self, csv_file_path, output_style="full"):
"write_gib": "Write (GiB/sec)",
"read_gib": "Read (GiB/sec)"
}
row_order = ["test_case", "start_time", "daos_commit", "oclass",
row_order = ["test_case", "start_time", "daos_commit", "provider", "oclass",
"num_servers", "num_clients", "ppc",
"write_gib", "read_gib",
"notes", "status"]
Expand Down Expand Up @@ -553,6 +580,7 @@ def process_result_file(self, file_path):

row["daos_commit"] = get_daos_commit(file_path, row["slurm_job_id"])
row["num_targets"] = get_num_targets(file_path, row["slurm_job_id"])
row["provider"] = get_provider(file_path, row["slurm_job_id"])
row["write_gib"] = format_float(wr_gib)
row["read_gib"] = format_float(rd_gib)
row["status"] = status.get_status_str()
Expand All @@ -576,6 +604,7 @@ def __init__(self, csv_file_path, output_style="full"):
"start_time": "Date",
"end_time": "End",
"daos_commit": "Commit",
"provider": "Provider",
"oclass": "Oclass",
"dir_oclass": "Dir Oclass",
"num_servers": "Num Servers",
Expand All @@ -594,7 +623,7 @@ def __init__(self, csv_file_path, output_style="full"):
"read_kops": "read(Kops/sec)",
"remove_kops": "remove(Kops/sec)"
}
row_order = ["test_case", "start_time", "daos_commit", "oclass",
row_order = ["test_case", "start_time", "daos_commit", "provider", "oclass",
"num_servers", "num_clients", "ppc", "create_kops",
"stat_kops", "read_kops", "remove_kops",
"notes", "status"]
Expand Down Expand Up @@ -649,6 +678,7 @@ def process_result_file(self, file_path):

row["daos_commit"] = get_daos_commit(file_path, row["slurm_job_id"])
row["num_targets"] = get_num_targets(file_path, row["slurm_job_id"])
row["provider"] = get_provider(file_path, row["slurm_job_id"])
row["status"] = status.get_status_str()
row["notes"] = status.get_notes_str()

Expand Down

0 comments on commit 810d6a7

Please sign in to comment.