mdce start throws ulimit errors

Question

Lachlan Simpson 2015-6-10

1
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/223199-mdce-start-throws-ulimit-errors

回答： Lachlan Simpson 2015-7-27

采纳的回答： Lachlan Simpson

在 MATLAB Online 中打开

When starting mdce, there are ulimit problems

user@mdcs-head:/usr/local/MATLAB/R2015a/toolbox/distcomp/bin$ sudo ./mdce start

WARNING: The mdce script detected that the number of processes allowed is limited by ulimit. Be sure that the limit of processes for the ROOT user (or the user running the mdce service) is set to either "unlimited" or at least 128 * W, where W is the maximum number of MDCS workers that will run on this machine.

I looked into raising ulimit and found this description on how to increase it.

I edited /etc/sysctl.conf to keep the changes permanent.

I rebooted the machine.

./mdce still throws the same error.

I re-installed everything, but it still throws the same error.

I re-read the error and was intrigued that it distinguished users. So I did another search and found that ulimit can be run per user :

user@mdcs-head:/usr/local/MATLAB/R2015a/toolbox/distcomp/bin$ ulimit 
unlimited
ubuntu@mdcs-head:/usr/local/MATLAB/R2015a/toolbox/distcomp/bin$ sudo -i
root@mdcs-head:~# ulimit
unlimited

Ok. So both users are unlimited. MDCS will run as root.

What's with the error?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Lachlan Simpson 2015-7-27

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/223199-mdce-start-throws-ulimit-errors#answer_187256

The actual problem was that ulimit is insufficient: need to use ulimit -a to see full values and respond accordingly.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

Answer 2

Ganesh Gaonkar 2015-6-12

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/223199-mdce-start-throws-ulimit-errors#answer_182428

The warning message sent by you shows that you are getting a communication failure.As a first step, validate your cluster profile and make sure that all the tests pass. If some tests fail, this would mean that this is an issue with your MDCS installation and configuration. You can do the validation tests as follows: Go to Parallel (under Home Tab in MATLAB Ribbon on top) ->Manage Cluster Profiles ->Validate.

If all validation tests pass, Refer the below links . These can help you with some debugging steps:

Link1

Link2