From 451c0f0510a001ad9179e57fd9c2f0761151fc18 Mon Sep 17 00:00:00 2001 From: 0xFugue <119708655+0xFugue@users.noreply.github.com> Date: Thu, 17 Aug 2023 21:34:29 +0530 Subject: [PATCH] MiB to MegaBytes renaming and the "divide" option (#132) * MiB to Mega Bytes renaming * remove fifo explicitly * modify the fifo name * added the per-container "divide"; configurable --- analysis-module/src/hproc.py | 84 +++++++++++++++++++++--------------- bash-utils/cleanup.sh | 2 +- run.sh | 4 +- 3 files changed, 53 insertions(+), 37 deletions(-) diff --git a/analysis-module/src/hproc.py b/analysis-module/src/hproc.py index e22d2cb..85fe92c 100644 --- a/analysis-module/src/hproc.py +++ b/analysis-module/src/hproc.py @@ -64,7 +64,7 @@ class Human2BytesConverter(metaclass=Singleton): # Base class for plots and common helpers class Plots(metaclass=Singleton): - def __init__(self, log_dir, oprefix, jf, to_plot, cfile): + def __init__(self, log_dir, oprefix, jf, to_plot, cfile, divide): self.log_dir, self.oprefix = log_dir, oprefix self.df, self.n, self.keys, self.cols = pd.DataFrame(), 0, [], [] self.col2title, self.col2units, self.key2nodes = {}, {}, {} @@ -73,7 +73,7 @@ class Plots(metaclass=Singleton): self.fig, self.axes = "", "" self.json_fname, self.G = jf, nx.empty_graph() self.to_plot, self.to_compare = to_plot, [] - self.run_summary, self.cfile = "", cfile + self.run_summary, self.cfile, self.divide, self.container_size = "", cfile, divide, 1.0 # waku log processing def compute_msg_settling_times(self): @@ -102,6 +102,7 @@ class Plots(metaclass=Singleton): def set_summary(self): with open(self.cfile, 'r') as f: # Load config file conf = json.load(f) + self.container_size = float(conf["gennet"]["container_size"]) minsize = int(conf["wls"]["min_packet_size"]/1024) maxsize = int(conf["wls"]["max_packet_size"]/1024) self.run_summary = (f'{conf["gennet"]["num_nodes"]}by' @@ -296,7 +297,11 @@ class Plots(metaclass=Singleton): col = self.to_compare[k] #self.axes[i,j].ticklabel_format(style='plain') self.axes[i,j].yaxis.grid(True) - pc = self.axes[i,j].violinplot(self.df[col], showmedians=True) + if self.divide and col != "CPUPerc": # Jordi's compare plots do not divide CPU + ddf = self.df[col]/self.container_size + pc = self.axes[i,j].violinplot(ddf, showmedians=True) + else: + pc = self.axes[i,j].violinplot(self.df[col], showmedians=True) self.axes[i,j].set_ylabel(self.col2units[col]) self.axes[i,j].set_title(self.col2title[col]) #for p in pc['bodies']: @@ -377,8 +382,8 @@ class Plots(metaclass=Singleton): # handle docker stats class DStats(Plots, metaclass=Singleton): - def __init__(self, log_dir, oprefix, jf, to_plot, cfile): - Plots.__init__(self, log_dir, oprefix, jf, to_plot, cfile) + def __init__(self, log_dir, oprefix, jf, to_plot, cfile, divide): + Plots.__init__(self, log_dir, oprefix, jf, to_plot, cfile, divide) self.dstats_fname = f'{log_dir}/dstats-data/docker-stats.out' self.kinspect_fname = f'{log_dir}/dstats-data/docker-kinspect.out' self.col2title = { "ContainerID" : "Docker ID", @@ -396,13 +401,13 @@ class DStats(Plots, metaclass=Singleton): self.col2units = { "ContainerID" : "ID", "ContainerName" : "Name", "CPUPerc" : "Percentage (%)", - "MemUse" : "MiB", - "MemTotal" : "MiB", + "MemUse" : "MegaBytes", + "MemTotal" : "MegaBytes", "MemPerc" : "Percentage (%)", - "NetRecv" : "MiB", - "NetSent" : "MiB", - "BlockR" : "MiB", - "BlockW" : "MiB", + "NetRecv" : "MegaBytes", + "NetSent" : "MegaBytes", + "BlockR" : "MegaBytes", + "BlockW" : "MegaBytes", "CPIDS" : "PIDS" } self.cols = ["CPUPerc", "MemUse","NetRecv", "NetSent", "BlockR", "BlockW"] @@ -425,12 +430,13 @@ class DStats(Plots, metaclass=Singleton): h2b, n = Human2BytesConverter(), len(self.keys) for percent in ["CPUPerc", "MemPerc"]: self.df[percent] = self.df[percent].str.replace('%','').astype(float) + # Normalise to MegaBytes for size in ["MemUse", "MemTotal"]: - self.df[size] = self.df[size].map(lambda x:h2b.convert(x.strip())/(1024*1024)) # MiB + self.df[size] = self.df[size].map(lambda x:h2b.convert(x.strip())/(1024*1024)) for size in ["NetRecv", "NetSent"]: - self.df[size] = self.df[size].map(lambda x:h2b.convert(x.strip())/(1024*1024)) # MiB + self.df[size] = self.df[size].map(lambda x:h2b.convert(x.strip())/(1024*1024)) for size in ["BlockR", "BlockW"]: - self.df[size] = self.df[size].map(lambda x:h2b.convert(x.strip())/(1024*1024)) # MiB + self.df[size] = self.df[size].map(lambda x:h2b.convert(x.strip())/(1024*1024)) self.df['Key'] = self.df['ContainerName'].map(lambda x: x.split("--")[0]) self.build_key2nodes() self.df['NodeName'] = self.df['Key'].map(lambda x: self.key2nodes[x][0]) @@ -451,8 +457,8 @@ class DStats(Plots, metaclass=Singleton): class HostProc(Plots, metaclass=Singleton): - def __init__(self, log_dir, oprefix, jf, to_plot, cfile): - Plots.__init__(self, log_dir, oprefix, jf, to_plot, cfile) + def __init__(self, log_dir, oprefix, jf, to_plot, cfile, divide): + Plots.__init__(self, log_dir, oprefix, jf, to_plot, cfile, divide) self.fname = f'{log_dir}/host-proc-data/docker-proc.out' self.kinspect_fname = f'{log_dir}/host-proc-data/docker-kinspect.out' self.col2title = { 'CPUPerc' : 'CPU Utilisation', @@ -474,22 +480,22 @@ class HostProc(Plots, metaclass=Singleton): 'BlockW' : 'Block Writes' } self.col2units = { 'CPUPerc' : '%', - 'VmPeak' : 'MiB', - 'MemUse' : 'MiB', - 'VmSize' : 'MiB', - 'VmRSS' : 'MiB', - 'VmData' : 'MiB', - 'VmStk' : 'MiB', - 'NetRecv' : 'MiB', + 'VmPeak' : 'MegaBytes', + 'MemUse' : 'MegaBytes', + 'VmSize' : 'MegaBytes', + 'VmRSS' : 'MegaBytes', + 'VmData' : 'MegaBytes', + 'VmStk' : 'MegaBytes', + 'NetRecv' : 'MegaBytes', 'NetRecvPkts' : 'Packets', - 'NetSent' : 'MiB', + 'NetSent' : 'MegaBytes', 'NetSentPkts' : 'Packets', - 'NetRX' : 'MiB', - 'NetWX' : 'MiB', - 'InOctets' : 'MiB', - 'OutOctets' : 'MiB', - 'BlockR' : 'MiB', - 'BlockW' : 'MiB' + 'NetRX' : 'MegaBytes', + 'NetWX' : 'MegaBytes', + 'InOctets' : 'MegaBytes', + 'OutOctets' : 'MegaBytes', + 'BlockR' : 'MegaBytes', + 'BlockW' : 'MegaBytes' } self.cols = ['CPUPerc', 'VmPeak', 'MemUse', 'VmSize', 'VmRSS', 'VmData', 'VmStk', 'RxBytes', 'RxPackets', 'TxBytes', 'TxPackets', 'NetRecv', 'NetSent', @@ -519,11 +525,11 @@ class HostProc(Plots, metaclass=Singleton): def post_process(self): #h2b = Human2BytesConverter() for size in ['VmPeak', 'VmSize','VmRSS', 'VmData','VmStk']: - self.df[size] = self.df[size].map(lambda x: x/1024) # MiBs + self.df[size] = self.df[size].map(lambda x: x/1024) # MegaBytes for size in ['NetRecv','NetSent']: - self.df[size] = self.df[size].map(lambda x: x/(1024*1024)) # MiBs + self.df[size] = self.df[size].map(lambda x: x/(1024*1024)) # MegaBytes for size in ['BlockR', 'BlockW']: - self.df[size] = self.df[size].map(lambda x: x/(1024*1024)) # MiBs + self.df[size] = self.df[size].map(lambda x: x/(1024*1024)) # MegaBytes #self.df['Key'] = self.df['ContainerName'].map(lambda x: x.split("--")[0]) self.df['Key'] = self.df['NodeName']#.map(lambda x: x.split("--")[0]) self.df['MemUse'] = self.df['VmPeak'] @@ -593,17 +599,21 @@ def cmd_helper(metric_infra, to_plot, agg, to_compare): def host_proc(ctx: typer.Context, log_dir: Path, # <- mandatory path out_prefix: str = typer.Option("output", help="Specify the prefix for the plot pdfs"), aggregate: bool = typer.Option(True, help="Specify whether to aggregate"), + divide : bool = typer.Option(False, + help="Specify if you want to divide by container size for compare plots"), config_file: str = typer.Option("", callback=_config_file_callback, is_eager=True, help="Set the input config file (JSON)")): if not path_ok(log_dir, True): sys.exit(0) + if not config_file : + config_file=f'{os.path.abspath(log_dir)}/config/config.json' # set the default config to_plot = ctx.default_map["to_plot"] if ctx.default_map and "to_plot" in ctx.default_map else [] jf = f'{os.path.abspath(log_dir)}/config/topology_generated/network_data.json' if os.path.exists("plots"): os.system('rm -rf plots') os.makedirs("plots") - host_proc = HostProc(log_dir, f'plots/{out_prefix}-host-proc', jf, to_plot, config_file) + host_proc = HostProc(log_dir, f'plots/{out_prefix}-host-proc', jf, to_plot, config_file, divide) cmd_helper(host_proc, to_plot, agg=aggregate, to_compare=["CPUPerc", "MemUse", "NetRecv", "NetSent", "BlockR", "BlockW"]) log.info(f'Done: {log_dir}') @@ -614,17 +624,21 @@ def host_proc(ctx: typer.Context, log_dir: Path, # <- mandatory path def dstats(ctx: typer.Context, log_dir: Path, # <- mandatory path out_prefix: str = typer.Option("output", help="Specify the prefix for the plot pdfs"), aggregate: bool = typer.Option(True, help="Specify whether to aggregate"), + divide : bool = typer.Option(False, + help="Specify if you want to divide by container size for compare plots"), config_file: str = typer.Option("", callback=_config_file_callback, is_eager=True, help="Set the input config file (JSON)")): if not path_ok(log_dir, True): sys.exit(0) + if not config_file : + config_file=f'{os.path.abspath(log_dir)}/config/config.json' # set the default config to_plot = ctx.default_map["to_plot"] if ctx.default_map and "to_plot" in ctx.default_map else [] jf = f'{os.path.abspath(log_dir)}/config/topology_generated/network_data.json' if os.path.exists("plots"): os.system('rm -rf plots') os.makedirs("plots") - dstats = DStats(log_dir, f'plots/{out_prefix}-dstats', jf, to_plot, config_file) + dstats = DStats(log_dir, f'plots/{out_prefix}-dstats', jf, to_plot, config_file, divide) cmd_helper(dstats, to_plot, agg=aggregate, to_compare=["CPUPerc", "MemUse", "NetRecv", "NetSent", "BlockR", "BlockW"]) log.info(f'Done: {log_dir}') diff --git a/bash-utils/cleanup.sh b/bash-utils/cleanup.sh index 0e00473..3b7d041 100755 --- a/bash-utils/cleanup.sh +++ b/bash-utils/cleanup.sh @@ -2,7 +2,7 @@ enclave_name=${1:-"wakurtosis"} # hardcoded files/fifo/folders rm -f ./kurtosisrun_log.txt -rm -f /tmp/hostproc-signal.fifo +rm -f /tmp/host-proc-signal.fifo rm -rf ./wakurtosis_logs ./config/topology_generated ./monitoring/host-proc/stats ./monitoring/dstats/stats monitoring/container-proc/cproc_metrics.json docker stop gennet cadvisor bootstrap_node dstats host-proc analysis > /dev/null 2>&1 diff --git a/run.sh b/run.sh index 0fdcc59..b6afda0 100755 --- a/run.sh +++ b/run.sh @@ -59,7 +59,7 @@ elif [ "$metrics_infra" = "host-proc" ]; then # HOST-PROC odir=./monitoring/host-proc/$stats_dir rclist=$odir/docker-rc-list.out mkdir $odir - mkfifo $signal_fifo + mkfifo $signal_fifo # get a fresh fifo for each run. chmod 0777 $signal_fifo # get the sudo sorted out in the main thread itself echo "host-proc: need sudo rights, please enter suitable credentials at the prompt" @@ -216,6 +216,8 @@ if [ "$metrics_infra" = "dstats" ]; then echo "dstats: copying the dstats data" cp -r ./monitoring/dstats/stats ${enclave_name}_logs/dstats-data elif [ "$metrics_infra" = "host-proc" ]; then + # do not reuse the fifo across runs + rm -f $signal_fifo echo "Copying the host-proc data" cp -r ./monitoring/host-proc/stats ${enclave_name}_logs/host-proc-data elif [ "$metrics_infra" = "container-proc" ]; then