Multi-Locale run errors

Users
1

Hi everyone,

I was working on Chapel program, after running an executable file in Spawnshell, it returned the following error.

Screenshot from 2021-11-18 14-02-58
Screenshot from 2021-11-18 14-02-581280×1024 243 KB

And after it, my remote nodes are not recognizable as you see as follows (Note: YangHost is my local machine)

Screenshot from 2021-11-18 14-00-11
Screenshot from 2021-11-18 14-00-111280×1024 213 KB

Any help would be appreciated,
Marjan

2

Hi Marjan —

The root cause looks similar to the previous error you had posted (which I don't think we had ever solved) related to Congestion at the destination endpoint. Did you have CHPL_COMM_DEBUG=1 set in your environment before compiling this program?

The mention of Microsoft.NETCore.App makes me realize: Are you running this from a Windows system? Are the nodes themselves WIndows systems as well? And in either case are you running this from Cygwin or using Microsoft's Bash Shell / Linux Subsystem?

Thanks,
-Brad

3

Hi Brad,
yes, I have set that variable, but I did not get any info about the actual problem. I am now thinking about migrating to Compute Canda Clusters. I have realized that these clusters support chapel. But to be sure, and if you have any information about these clusters, do I need to have any specific configurations?
here is the only doc I have found -->
https://docs.computecanada.ca/wiki/Chapel
Regarding .NET app, no I am working on Linux systems, and I am calling Spanwshell to run dotnet commands whenever I need to have access to the SQLite database from my Chapel program.

Thank you,
Marjan

4

Hi Marjan —

I have not run on Compute Canada, but some of our users and team members have, so know that it is possible. I don't know whether there are specific settings or invocations to know about but am tagging @e-kayrakli and @simonBourgaultCote who may be able to provide more information.

Thanks for the clarification about the Microsoft text, that makes sense. In a UNIX setting, I'm not aware of anything Chapel or GASNet could do to mess up a system offhand. I'll check if the GASNet team has any further thoughts about these errors you're getting.

-Brad

5

I do really appreciate all of these. Thank you, Brad

6

@MarjanAsgari — Before we give up on your local system, would you:

  • make sure CHPL_COMM_DEBUG=1 is still set
  • compile and run one of our example programs:
    $ chpl examples/hello6-taskpar-dist.chpl  # if you're working from a git clone, this is test/release/examples/hello6-taskpar-dist.chpl
$ ./hello6-taskpar-dist -nl 3   # or however many locales you've been trying to run on
  • send me the full output (preferably as cut-and-pasted text rather than a screenshot, if possible)

Also, is there any chance that these nodes you're using are running any sort of filters protecting against (D)DoS attacks?

Thanks,
-Brad

7

Hi Brad, Sure, I’ll try it this suggestion tomorrow and will send the full output to you. Thank you