maurorappa

package main

import (
	"bufio"
	"bytes"
	"flag"
	"fmt"
	"github.com/BurntSushi/toml"
	"github.com/kavehmz/prime"
	"io/ioutil"

maurorappa

FROM gcc:11.2.0 as build
RUN apt update && apt install -y p7zip fio python3 g++ ipmitool groff-base dmidecode perl vim python3-pymongo && mkdir /hwtest
COPY . hwtest/
RUN ln -s /usr/local/bin/gcc /usr/bin/gcc && mkdir /etc_phy /data
RUN cd /hwtest && g++ -O3 -DNDEBUG c2clat.cpp -o c2clat -pthread
RUN cd /hwtest && tar xf  memtester-4.6.0.tar.gz && tar xf primegen-0.97.tar.gz && cp /hwtest/memtester.c  /hwtest/memtester-4.6.0/ && cp /hwtest/primespeed.c /hwtest/primegen-0.97/
RUN cd /hwtest/primegen-0.97/ && make && cp /hwtest/primegen-0.97/primespeed /bin/primespeed
RUN cd /hwtest/memtester-4.6.0/ && make && cp /hwtest/memtester-4.6.0/memtester /bin/memtester
RUN git clone https://github.com/cruvolo/ramspeed-smp.git && cd ramspeed-smp && echo |bash build.sh && cp ramsmp /bin/ramsmp
RUN git clone https://github.com/eembc/coremark-pro.git coremark-pro && cd coremark-pro && make TARGET=linux64 XCMD='-c4' all

maurorappa

    microservices in containers without overengineering

    all communications by BUS (easy to monitor everything, easy to add new sites, easy to test, easy to implement business logic )

    all messages in json with schema (allows easy parsing/storage/replay and multiple service versions to coexists)

    central db and Redis NOT in the cluster (shared storage as single point of truth, no cache sync problem)

    api routing with nginx or openresty (http routing inside the gateway, no application gw)

maurorappa

How many write syscalls?

# perf stat -e 'syscalls:sys_enter_writev'  -p 546925
^C
 Performance counter stats for process id '546925':

                 9      syscalls:sys_enter_writev

       3.400728456 seconds time elapsed

maurorappa

The kernel creates a thread for every md raid configured:
# ps -ef|grep raid
root         912       2  0 15:24 ?        00:00:00 [raid5wq]
root        3383       2  0 15:24 ?        00:00:00 [md0_raid5]
root        3402       2  0 15:24 ?        00:00:03 [md1_raid5]


Being kernel threads, its management is limited, they cannot be pinned to CPU therefore they can move accross all available CPUs, they can get hammered by IRQ, they might have a 'busy' sibling…
Probably dedicated CPU will provide better performance (testing will follow!).

maurorappa

 bpftrace -e 'tracepoint:syscalls:sys_enter_exec*{ printf("pid: %d, comm: %s, args: ", pid, comm); join(args->argv); }

maurorappa

#sqlite3 /var/lib/grafana/grafana.db
 
SQLite version 3.7.17 
Enter ".help" for instructions
Enter SQL statements terminated with a ";"
sqlite> .output dump.sql
sqlite> .dump dashboard
sqlite> .exit

# tail dump.sql

maurorappa

# tshark -r /root/md.pcap -q -n -z dests,tree
Running as user "root" and group "root". This could be dangerous.

=======================================================================================================================================
IPv4 Statistics/Destinations and Ports:
Topic / Item            Count         Average       Min val       Max val       Rate (ms)     Percent       Burst rate    Burst start
---------------------------------------------------------------------------------------------------------------------------------------
Destinations and Ports  127892                                                  12.7895       100%          49.6000       0.075
 222.10.9.3             127892                                                  12.7895       100.00%       49.6000       0.075
  UDP                   127892                                                  12.7895       100.00%       49.6000       0.075

maurorappa

https://www.ele.uri.edu/research/hpcl/2012/SBAC.pdf


how it compiled:

objdump -s --section .gnu.build.attributes /usr/lib64/liblzo2.so.2.0.0

or check the spec file! face with tongue 

maurorappa

from man adjtimex :

Linux  uses David L. Mills' clock adjustment algorithm (see RFC 1305).  The system call adjtimex() reads and optionally sets adjustment parameters for this algorithm.

on Centos7 , you can use systemtap to monitor who calls this syscall.
you need to install all kernel development and debuginfo packages and only then you can run this:

    #cat clock.stp
    probe kernel.function("sys_adjtimex") { printf("time adjusted by %s(%d)\n",execname(), pid()); }