在 2020/8/26 上午9:19, Daniel Jordan 写道:
> On Tue, Aug 25, 2020 at 11:26:58AM +0800, Alex Shi wrote:
>> 在 2020/8/25 上午9:56, Daniel Jordan 写道:
>>> Alex, do you have a pointer to the modified readtwice case?
>>
>> Sorry, no. my developer machine crashed, so I lost case my container and modified
>> case. I am struggling to get my container back from a account problematic repository. 
>>
>> But some testing scripts is here, generally, the original readtwice case will
>> run each of threads on each of cpus. The new case will run one container on each cpus,
>> and just run one readtwice thead in each of containers.
> 
> Ok, what you've sent so far gives me an idea of what you did.  My readtwice
> changes were similar, except I used the cgroup interface directly instead of
> docker and shared a filesystem between all the cgroups whereas it looks like
> you had one per memcg.  30 second runs on 5.9-rc2 and v18 gave 11% more data
> read with v18.  This was using 16 cgroups (32 dd tasks) on a 40 CPU, 2 socket
> machine.

I clean up my testing and make it reproducable by a Dockerfile and a case patch which
attached. 
User can build a container from the file, and then do testing like following:

#start some testing containers
for ((i=0; i< 80; i++)); do docker run --privileged=true --rm lrulock bash -c " sleep 20000" & done

#do testing evn setup 
for i in `docker ps | sed '1 d' | awk '{print $1 }'` ;do docker exec --privileged=true -it $i bash -c "cd vm-scalability/; bash -x ./case-lru-file-readtwice m"& done

#kick testing
for i in `docker ps | sed '1 d' | awk '{print $1 }'` ;do docker exec --privileged=true -it $i bash -c "cd vm-scalability/; bash -x ./case-lru-file-readtwice r"& done

#show result
for i in `docker ps | sed '1 d' | awk '{print $1 }'` ;do echo === $i ===; docker exec $i bash -c 'cat /tmp/vm-scalability-tmp/dd-output-* '  & done | grep MB | awk 'BEGIN {a=0;} { a+=$10 } END {print NR, a/(NR)}'

This time, on a 2P * 20 core * 2 HT machine,
This readtwice performance is 252% compare to v5.9-rc2 kernel. A good surprise!

> 
>>> Even better would be a description of the problem you're having in production
>>> with lru_lock.  We might be able to create at least a simulation of it to show
>>> what the expected improvement of your real workload is.
>>
>> we are using thousands memcgs in a machine, but as a simulation, I guess above case
>> could be helpful to show the problem.
> 
> Using thousands of memcgs to do what?  Any particulars about the type of
> workload?  Surely it's more complicated than page cache reads :)

Yes, the workload are quit different on different business, some use cpu a lot, some use
memory a lot, and some are may mixed. For containers number, that are also quit various
from tens to hundreds to thousands.

> 
>>> I ran a few benchmarks on v17 last week (sysbench oltp readonly, kerndevel from
>>> mmtests, a memcg-ized version of the readtwice case I cooked up) and then today
>>> discovered there's a chance I wasn't running the right kernels, so I'm redoing
>>> them on v18.
> 
> Neither kernel compile nor git checkout in the root cgroup changed much, just
> 0.31% slower on elapsed time for the compile, so no significant regressions
> there.  Now for sysbench again.
> 

Thanks a lot for testing report!
Alex