Re: [KVM-AUTOTEST][PATCH] timedrift support

[Yaniv Kaul <ykaul@redhat.com>]

On 5/11/2009 1:40 PM, Bear Yang wrote:
> Hello.
> I have modified my script according Marcelo's suggestion. and resubmit 
> my script to you all. :)
>
> Marcelo, Seems except you, no one care my script. I  still want to say 
> any suggestion  on my script would be greatly appreciated.
>
> Thanks.
>
> Bear
>
> Marcelo Tosatti wrote:
>> Bear,
>>
>> Some comments below.
>>
>> On Wed, May 06, 2009 at 12:02:21PM +0800, Bear Yang wrote:
>>> Hello everyone,
>>>
>>> I like to submit patch to add a new function for 'time drift check' 
>>> for  guest running on KVM.
>>>
>>> The TimeDrift design logic is below:
>>> 1. Set the host as the NTP server
>>> 2. Guest only sync it's clock with host *once* when it booted up.
>>> * if the offset value of ntpdate large than 1 sec, the guest will 
>>> sync  the clock with host.
>>> * if the offset value of ntpdate less than 1 sec, the guest doesn't 
>>> need  sync it's clock with host.
I don't see the value of doing all the NTP stuff above. It poses another 
requirement on the host.
>>>
>>> 3. Then the cpu stress testing will running on guest.
>>> * a C program will give the real load to guest cpu
Isn't there any available program that can perform a load on the guest 
CPU? Moreover, I believe we want different kinds of loads to be tested 
(mixed with I/O, for example).
>>> 4.when the cpustress testing finished. running the commandline 
>>> <ntpdate  -q host-ip> totally 20 times on guest to query the time 
>>> from host and  judge whether the guest clock has drift or not.
>>>
>>> The details of my patch is attached.
>>>
>>> thanks.
>>>
>>> Bear.

I might be missing something, but generally, what I feel should be the 
test is:
1. Load the guest (AND the host - configurable!)
2. Run a program for X minutes on the guest - as measured by the guest 
(say 3 minutes).
3. When it exits, check that indeed Xhost = Xguest (in some cases we've 
seen here: 3:15 on the host, vs. 3:00 on the guest).

Looks like a much simpler plan than yours. And should also be pretty 
portable to Windows.

Y.
>>
>>> diff -urN kvm_runtest_2.bak/cpu_stress.c kvm_runtest_2/cpu_stress.c
>>> --- kvm_runtest_2.bak/cpu_stress.c    1969-12-31 19:00:00.000000000 
>>> -0500
>>> +++ kvm_runtest_2/cpu_stress.c    2009-05-05 22:35:34.000000000 -0400
>>> @@ -0,0 +1,61 @@
>>> +#define _GNU_SOURCE
>>> +#include <stdio.h>
>>> +#include <pthread.h>
>>> +#include <sched.h>
>>> +#include <stdlib.h>
>>> +#include <fcntl.h>
>>> +#include <math.h>
>>> +#include <unistd.h>
>>> +
>>> +#define MAX_CPUS 256
>>> +#define BUFFSIZE 1024
>>> +
>>> +
>>> +void worker_child(int cpu)
>>> +{
>>> +    int cur_freq;
>>> +    int min_freq;
>>> +    int max_freq;
>>> +    int last_freq;
>>> +    cpu_set_t mask;
>>> +    int i;
>>> +    double x;
>>> +        int d = 0;
>>> +    /*
>>> +     * bind this thread to the specified cpu +     */
>>> +    CPU_ZERO(&mask);
>>> +    CPU_SET(cpu, &mask);
>>> +    sched_setaffinity(0, CPU_SETSIZE, &mask);
>>> +
>>> +    while (d++ != 500000) {
>>> +            for (i=0; i<100000; i++)
>>> +                x = sqrt(x);
>>> +    }
>>> +
>>> +    _exit(0);
>>> +
>>> +}
>>> +
>>> +
>>> +main() {
>>> +    cpu_set_t mask;
>>> +    int i;
>>> +    int code;
>>> +
>>> +    if (sched_getaffinity(0, CPU_SETSIZE, &mask) < 0){
>>> +        perror ("sched_getaffinity");
>>> +        exit(1);
>>> +    }
>>> +
>>> +    for (i=0; i<CPU_SETSIZE; i++)
>>> +        if (CPU_ISSET(i, &mask)){
>>> +            printf ("CPU%d\n",i);
>>> +            if (fork() == 0)
>>> +                worker_child(i);
>>> +        }
>>> +
>>> +
>>> +    wait(&code);
>>> +    exit (WEXITSTATUS(code));
>>> +}
>>> diff -urN kvm_runtest_2.bak/kvm_runtest_2.py 
>>> kvm_runtest_2/kvm_runtest_2.py
>>> --- kvm_runtest_2.bak/kvm_runtest_2.py    2009-04-29 
>>> 06:17:29.000000000 -0400
>>> +++ kvm_runtest_2/kvm_runtest_2.py    2009-04-29 08:06:32.000000000 
>>> -0400
>>> @@ -36,6 +36,8 @@
>>>                  "autotest":     test_routine("kvm_tests",           
>>> "run_autotest"),
>>>                  "kvm_install":  test_routine("kvm_install",         
>>> "run_kvm_install"),
>>>                  "linux_s3":     test_routine("kvm_tests",           
>>> "run_linux_s3"),
>>> +                "ntp_server_setup": test_routine("kvm_tests",       
>>> "run_ntp_server_setup"),
>>> +                "timedrift":    test_routine("kvm_tests",           
>>> "run_timedrift"),
>>>                  }
>>>
>>>          # Make it possible to import modules from the test's bindir
>>> diff -urN kvm_runtest_2.bak/kvm_tests.cfg.sample 
>>> kvm_runtest_2/kvm_tests.cfg.sample
>>> --- kvm_runtest_2.bak/kvm_tests.cfg.sample    2009-04-29 
>>> 06:17:29.000000000 -0400
>>> +++ kvm_runtest_2/kvm_tests.cfg.sample    2009-04-29 
>>> 08:09:36.000000000 -0400
>>> @@ -81,6 +81,10 @@
>>>      - linux_s3:      install setup
>>>          type = linux_s3
>>>
>>> +    - ntp_server_setup:
>>> +        type = ntp_server_setup
>>> +    - timedrift:      ntp_server_setup
>>> +        type = timedrift
>>>  # NICs
>>>  variants:
>>>      - @rtl8139:
>>> diff -urN kvm_runtest_2.bak/kvm_tests.py kvm_runtest_2/kvm_tests.py
>>> --- kvm_runtest_2.bak/kvm_tests.py    2009-04-29 06:17:29.000000000 
>>> -0400
>>> +++ kvm_runtest_2/kvm_tests.py    2009-05-05 23:45:57.000000000 -0400
>>> @@ -394,3 +394,235 @@
>>>      kvm_log.info("VM resumed after S3")
>>>
>>>      session.close()
>>> +
>>> +def run_ntp_server_setup(test, params, env):
>>> +    +    """NTP server configuration and related network file 
>>> modification
>>> +    """
>>> +    kvm_log.debug("run ntp server setup")
>>> +    status = 1
>>> +    # stop firewall for NTP server if it is running.
>>> +    status = os.system("/etc/init.d/iptables status")
>>> +    if status == 0:
>>> +        os.system("/etc/init.d/iptables stop")
>>> +        status = 1
>>> +
>>> +    # prevent dhcp client modify the ntp.conf
>>> +    kvm_log.info("prevent dhcp client modify the ntp.conf")
>>> +
>>> +    config_file = "/etc/sysconfig/network"
>>> +    network_file = open("/etc/sysconfig/network", "a")
>>> +    string = "PEERNTP=no"
>>> +
>>> +    if os.system("grep %s %s" % (string, config_file)):
>>> +        network_file.writelines(str(string)+'\n')
>>> +    +    network_file.close()
>>> +  +    # start ntp server on host
>>> +    kvm_log.info("backup ntp config file")
>>> +
>>> +    ntp_filename = os.path.expanduser("/etc/ntp.conf")
>>> +    # backup ntp config file
>>> +    backup_bootloader_filename = ntp_filename + "_backup"
>>> +    if os.path.exists(ntp_filename):
>>> +        os.rename(ntp_filename, backup_bootloader_filename)
>>> +    +    status = os.system("/etc/init.d/ntpd status")
>>> +    if status == 0:
>>> +        os.system("/etc/init.d/ntpd stop")
>>> +        status = 1    +
>>> +    kvm_log.info("start ntp server on host")
>>> +
>>> +    ntp_cmd = '''
>>> +        echo "restrict default kod nomodify notrap nopeer noquery" 
>>> >> /etc/ntp.conf;\
>>> +        echo "restrict 127.0.0.1" >> /etc/ntp.conf;\
>>> +        echo "driftfile /var/lib/ntp/drift" >> /etc/ntp.conf;\
>>> +        echo "keys /etc/ntp/keys" >> /etc/ntp.conf;\
>>> +        echo "server 127.127.1.0" >> /etc/ntp.conf;\
>>> +        echo "fudge 127.127.1.0 stratum 1" >> /etc/ntp.conf;\
>>> +        service ntpd start;
>>> +        '''
>>
>> I think it would be better to copy /etc/ntp.conf to a temporary file,
>> modify that, and start ntpd with the -c option.
>>
>> After the test is finished, restart ntpd with the default config (if it
>> was running) via service ntpd restart.
>>
>> Also I don't see whether your script reports the content of
>>
>> /sys/devices/system/clocksource/clocksource0/current_clocksource
>>
>> On the guest? Its important that information is displayed on the test
>> report.
>>
>> Looks fine to me other than that, but the kvm-autotest guys probably
>> have more comments.
>>
>> Thanks
>> -- 
>> To unsubscribe from this list: send the line "unsubscribe kvm" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>