Re: [PATCH v2 8/8] simplebench/bench_block_job: drop caches before test run

[John Snow <jsnow@redhat.com>]

On 3/5/21 11:50 AM, Vladimir Sementsov-Ogievskiy wrote:
> 05.03.2021 19:30, John Snow wrote:
>> On 3/5/21 4:11 AM, Vladimir Sementsov-Ogievskiy wrote:
>>> 05.03.2021 04:30, John Snow wrote:
>>>> On 3/4/21 5:17 AM, Vladimir Sementsov-Ogievskiy wrote:
>>>>> It probably may improve reliability of results when testing in cached
>>>>> mode.
>>>>>
>>>>> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
>>>>> ---
>>>>>   scripts/simplebench/bench_block_job.py | 2 ++
>>>>>   1 file changed, 2 insertions(+)
>>>>>
>>>>> diff --git a/scripts/simplebench/bench_block_job.py 
>>>>> b/scripts/simplebench/bench_block_job.py
>>>>> index 4f03c12169..fa45ad2655 100755
>>>>> --- a/scripts/simplebench/bench_block_job.py
>>>>> +++ b/scripts/simplebench/bench_block_job.py
>>>>> @@ -53,6 +53,8 @@ def bench_block_job(cmd, cmd_args, qemu_args):
>>>>>           return {'error': 'qemu failed: ' + str(vm.get_log())}
>>>>>       try:
>>>>> +        subprocess.run('sync; echo 3 > /proc/sys/vm/drop_caches', 
>>>>> shell=True,
>>>>> +                       check=True)
>>>>>           res = vm.qmp(cmd, **cmd_args)
>>>>>           if res != {'return': {}}:
>>>>>               vm.shutdown()
>>>>>
>>>>
>>>> Worth adding a conditional to allow "hot" or "cold" runs? nah?
>>>>
>>>
>>> You mean, make this addition optional? Make sense
>>>
>>>
>>
>> I was thinking (along the lines of allowing both old and new behavior, 
>> in case anyone except you used these scripts) of this sort of thing:
>>
>> def bench_block_job(cmd, cmd_args, qemu_args, drop_cache=True): ...
>>
>> I don't insist on it; I was just earnestly wondering if it had any 
>> utility. If it doesn't, don't respin on my account.
>>
> 
> Ok, thanks a lot for reviewing! Still, I think, I'll resend
> 

Thanks for sharing your benchmarking scripts :)

--js