added gitlab testing infrastructure to the test

[kosinovsky]

added gitlab testing infrastructure to the test

[adammoody]

Thanks, @kosinovsky . This was another test I wrote initially to do interactive debugging with TotalView. It's good for that, but let's improve it in a few ways for our automated testing. Would you please update this to do the following:

1. after the flush, delete the source files
2. during the fetch, check that each process gets back the set of filenames it expects
3. write something unique to the file for each process before the flush, and then after the fetch read the files in and verify that the contents are exactly as the process expects

Thanks!

[kosinovsky]

Sounds good, Adam. I will do that. Thank you, Greg From: Adam Moody <[email protected]> Sent: Wednesday, June 26, 2019 3:27 PM To: ECP-VeloC/filo <[email protected]> Cc: Kosinovsky, Greg <[email protected]>; Mention <[email protected]> Subject: Re: [ECP-VeloC/filo] added gitlab testing infrastructure to the test (#3) Thanks, @kosinovsky<https://github.com/kosinovsky> . This was another test I wrote initially to do interactive debugging with TotalView. It's good for that, but let's improve it in a few ways for our automated testing. Would you please update this to do the following: 1. after the flush, delete the source files 2. during the fetch, check that each process gets back the set of filenames it expects 3. write something unique to the file for each process before the flush, and then after the fetch read the files in and verify that the contents are exactly as the process expects Thanks! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#3?email_source=notifications&email_token=ACBP7IYVXQSEYCR4Y2M5FTTP4PUJ3A5CNFSM4H3A6WPKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODYU7WGQ#issuecomment-506067738>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ACBP7I2CZDRKZPUWYKVYNGLP4PUJ3ANCNFSM4H3A6WPA>.

[adammoody]

In addition to the test to check filenames and file contents on a normal flush/fetch mentioned above, let's add a couple more tests.

Let's build a test in which we execute the async functions of filo.

Let's build a test in which we execute a flush, delete one of the destination files for one of the ranks, then execute a fetch. We should see that all processes detect that one process failed to read one of its files. Because there is an allreduce across all procs, the return code in Filo_Fetch should be an error on all process, even on those processes whose files were not deleted.

[kosinovsky]

Hi Adam. Just wondering (regarding the point #1). After flush, why did you use “unlink” instead of “remove” in the first place (I presume that “remove” is what you mean by “delete”). In this case, since we don’t have multiple threads or processes accessing the same file, wouldn’t “unlink” and “remove” behave identically anyway? Thank you, Greg From: Adam Moody <[email protected]> Sent: Wednesday, June 26, 2019 3:27 PM To: ECP-VeloC/filo <[email protected]> Cc: Kosinovsky, Greg <[email protected]>; Mention <[email protected]> Subject: Re: [ECP-VeloC/filo] added gitlab testing infrastructure to the test (#3) Thanks, @kosinovsky<https://github.com/kosinovsky> . This was another test I wrote initially to do interactive debugging with TotalView. It's good for that, but let's improve it in a few ways for our automated testing. Would you please update this to do the following: 1. after the flush, delete the source files 2. during the fetch, check that each process gets back the set of filenames it expects 3. write something unique to the file for each process before the flush, and then after the fetch read the files in and verify that the contents are exactly as the process expects Thanks! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#3?email_source=notifications&email_token=ACBP7IYVXQSEYCR4Y2M5FTTP4PUJ3A5CNFSM4H3A6WPKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODYU7WGQ#issuecomment-506067738>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ACBP7I2CZDRKZPUWYKVYNGLP4PUJ3ANCNFSM4H3A6WPA>.

[adammoody]

@kosinovsky , yes "remove" would be fine as well.

[kosinovsky]

Hi Adam. Sorry, I did not stat my question clearly. You asked me to delete the source files after flush. However, they are already being deleted with the “unlink” command. I thought you meant you wanted me to replace “unlink” with "remove" – and I was wondering why. Thank you, Greg From: Adam Moody <[email protected]> Sent: Tuesday, July 16, 2019 11:03 PM To: ECP-VeloC/filo <[email protected]> Cc: Kosinovsky, Greg <[email protected]>; Mention <[email protected]> Subject: Re: [ECP-VeloC/filo] added gitlab testing infrastructure to the test (#3) @kosinovsky<https://github.com/kosinovsky> , yes "remove" would be fine as well. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#3?email_source=notifications&email_token=ACBP7IZMTURBBP3762SLEJTP72YZVA5CNFSM4H3A6WPKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2DEF6Y#issuecomment-512115451>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ACBP7I3KUUMJXL4HWECWJQ3P72YZVANCNFSM4H3A6WPA>.

[adammoody]

Oh, I see. I forgot that we were already deleting the source files. Yes, so step #1 of 3 is already done then.

[kosinovsky]

Thank you 😊 From: Adam Moody <[email protected]> Sent: Wednesday, July 17, 2019 11:17 AM To: ECP-VeloC/filo <[email protected]> Cc: Kosinovsky, Greg <[email protected]>; Mention <[email protected]> Subject: Re: [ECP-VeloC/filo] added gitlab testing infrastructure to the test (#3) Oh, I see. I forgot that we were already deleting the source files. Yes, so step #1<#1> of 3 is already done then. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#3?email_source=notifications&email_token=ACBP7I2O3H4EAQ5YMWRKOX3P75OZ7A5CNFSM4H3A6WPKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2GENXY#issuecomment-512509663>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ACBP7I5DIG7FLNGKCRSQMULP75OZ7ANCNFSM4H3A6WPA>.

[kosinovsky]

Hi Adam. Sorry, I just noticed this email from last week. Which “async functions” are you referring to below? Thank you, Greg From: Adam Moody <[email protected]> Sent: Tuesday, July 9, 2019 5:15 PM To: ECP-VeloC/filo <[email protected]> Cc: Kosinovsky, Greg <[email protected]>; Mention <[email protected]> Subject: Re: [ECP-VeloC/filo] added gitlab testing infrastructure to the test (#3) In addition to the test to check filenames and file contents on a normal flush/fetch mentioned above, let's add a couple more tests. Let's build a test in which we execute the async functions of filo. Let's build a test in which we execute a flush, delete one of the destination files for one of the ranks, then execute a fetch. We should see that all processes detect that one process failed to read one of its files. Because there is an allreduce across all procs, the return code in Filo_Fetch should be an error on all process, even on those processes whose files were not deleted. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#3?email_source=notifications&email_token=ACBP7I32TUXKOYKYHDTID3TP6USYPA5CNFSM4H3A6WPKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZR4WAY#issuecomment-509856515>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ACBP7I6OHYQKOENV65YWHV3P6USYPANCNFSM4H3A6WPA>.

[kosinovsky]

Thank you, Adam. Pushed the changes. Greg From: Adam Moody <[email protected]> Sent: Thursday, July 25, 2019 10:38 AM To: ECP-VeloC/filo <[email protected]> Cc: Kosinovsky, Greg <[email protected]>; Mention <[email protected]> Subject: Re: [ECP-VeloC/filo] added gitlab testing infrastructure to the test (#3) @adammoody commented on this pull request.

________________________________ In test/test_filo_async.c<#3 (comment)>:

+ char** dst_filelist;

+ /* src base path is still NULL (consistent with Filo_Flush), but the dest base path is /dev/shm*/ + rc = Filo_Fetch("mapfile", NULL, "/dev/shm", &num_files, &src_filelist, &dst_filelist, MPI_COMM_WORLD); + + /* free file list returned by fetch */ + int i; + for (i = 0; i < num_files; i++) { + //in file name, rank precedes ".out" suffix + int rank_from_file_name = *((strstr(dst_filelist[i], ".out"))-1) - '0'; + //assertain that the filename with consistant process marker was passed through flush/fetch + if(rank != rank_from_file_name){ + rc = TEST_FAIL; + printf("rank = %d, rank_from_file_name = %d\n", rank, rank_from_file_name); + } + //assertain that the file content is consistent with the process + FILE *file = fopen(filename, "r"); Same here, let's open the file using the name in dst_filelist, and close when done reading. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#3?email_source=notifications&email_token=ACBP7IZQ7G4KBKGMSIPELITQBHQHDA5CNFSM4H3A6WPKYY3PNVWWK3TUL52HS4DFWFIHK3DMKJSXC5LFON2FEZLWNFSXPKTDN5WW2ZLOORPWSZGOB7TLVFA#pullrequestreview-266779284>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ACBP7I4D6YNUN5SPOEL6WLTQBHQHDANCNFSM4H3A6WPA>.

[adammoody]

Thanks, @kosinovsky ! Looks good to me.

[CamStan]

I'll take a look at this and add a travis file and then it should be good.

[kosinovsky]

Thank you From: Cameron Stanavige <[email protected]> Sent: Thursday, July 25, 2019 4:46 PM To: ECP-VeloC/filo <[email protected]> Cc: Kosinovsky, Greg <[email protected]>; Mention <[email protected]> Subject: Re: [ECP-VeloC/filo] added gitlab testing infrastructure to the test (#3) I'll take a look at this and add a travis file and then it should be good. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#3?email_source=notifications&email_token=ACBP7I7YSE7YD7TRGFYKCELQBI3MFA5CNFSM4H3A6WPKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD23CSLA#issuecomment-515254572>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ACBP7IY47OVB3UD747DTIT3QBI3MFANCNFSM4H3A6WPA>.