Closed grondo closed 1 year ago
When updating the project Dockerfile to use the el8 image instead of centos8, it appeared that the v5.0.x openmpi branch now requires openmpix-4.2.
el8
centos8
*** Configuring PMIx checking for pmix pkg-config name... pmix checking if pmix pkg-config module exists... yes checking for pmix pkg-config cflags... checking for pmix pkg-config ldflags... checking for pmix pkg-config static ldflags... checking for pmix pkg-config libs... -lpmix -lhwloc -levent -lz checking for pmix pkg-config static libs... -lpmix -lhwloc -lm -ldl -lpthread -levent -lcrypto -lz checking pmix.h usability... yes checking pmix.h presence... yes checking for pmix.h... yes checking for PMIx_Init... yes checking if pmix requires libnl v1 or v3... none checking if external PMIx version is 4.1.2 or greater... no configure: error: External PMIx requested but not found.
Updating to that version, however triggers failures in pmix.nodeid.
pmix.nodeid
Changes to Dockerfile to reproduce:
diff --git a/src/test/docker/el8/Dockerfile b/src/test/docker/el8/Dockerfile index cae69e1..74a6165 100644 --- a/src/test/docker/el8/Dockerfile +++ b/src/test/docker/el8/Dockerfile @@ -1,9 +1,9 @@ -FROM fluxrm/flux-core:centos8 +FROM fluxrm/flux-core:el8 ARG USER=fluxuser ARG UID=1000 ARG OMPI_BRANCH=v5.0.x -ARG OPENPMIX_BRANCH=v4.1.1rc2 +ARG OPENPMIX_BRANCH=v4.2 RUN \ if test "$USER" != "fluxuser"; then \
expecting success: cat >2n3p.pmix.nodeid.exp <<-EOT && 0: 0 1: 0 2: 1 EOT run_timeout 30 flux mini run -N2 -n3 \ ${GETKEY} --label-io pmix.nodeid \ | sort -n >2n3p.pmix.nodeid.out && test_cmp 2n3p.pmix.nodeid.exp 2n3p.pmix.nodeid.out f4nALqDZ.2: PMIx_Get pmix.nodeid: NOT-FOUND f4nALqDZ.0: PMIx_Get pmix.nodeid: NOT-FOUND f4nALqDZ.1: PMIx_Get pmix.nodeid: NOT-FOUND flux-job: task(s) exited with exit code 1 --- 2n3p.pmix.nodeid.exp 2022-07-01 02:55:55.177460840 +0000 +++ 2n3p.pmix.nodeid.out 2022-07-01 02:55:55.177460840 +0000 @@ -1,3 +0,0 @@ -0: 0 -1: 0 -2: 1 not ok 18 - 2n3p pmix.nodeid is set correctly # # cat >2n3p.pmix.nodeid.exp <<-EOT && # 0: 0 # 1: 0 # 2: 1 # EOT # run_timeout 30 flux mini run -N2 -n3 \ # ${GETKEY} --label-io pmix.nodeid \ # | sort -n >2n3p.pmix.nodeid.out && # test_cmp 2n3p.pmix.nodeid.exp 2n3p.pmix.nodeid.out # expecting success: cat >pmix.nodeid.exp <<-EOT && 0: 0 1: 0 2: 1 3: 1 EOT run_timeout 30 flux mini run -N2 -n4 \ ${GETKEY} --label-io pmix.nodeid \ | sort -n >pmix.nodeid.out && test_cmp pmix.nodeid.exp pmix.nodeid.out f52wFg6s.0: PMIx_Get pmix.nodeid: NOT-FOUND f52wFg6s.2: PMIx_Get pmix.nodeid: NOT-FOUND f52wFg6s.3: PMIx_Get pmix.nodeid: NOT-FOUND f52wFg6s.1: PMIx_Get pmix.nodeid: NOT-FOUND flux-job: task(s) exited with exit code 1 --- pmix.nodeid.exp 2022-07-01 02:55:55.737465873 +0000 +++ pmix.nodeid.out 2022-07-01 02:55:55.741465909 +0000 @@ -1,4 +0,0 @@ -0: 0 -1: 0 -2: 1 -3: 1 not ok 19 - 2n4p pmix.nodeid is set correctly # # cat >pmix.nodeid.exp <<-EOT && # 0: 0 # 1: 0 # 2: 1 # 3: 1 # EOT # run_timeout 30 flux mini run -N2 -n4 \ # ${GETKEY} --label-io pmix.nodeid \ # | sort -n >pmix.nodeid.out && # test_cmp pmix.nodeid.exp pmix.nodeid.out #
When updating the project Dockerfile to use the
el8
image instead ofcentos8
, it appeared that the v5.0.x openmpi branch now requires openmpix-4.2.Updating to that version, however triggers failures in
pmix.nodeid
.Changes to Dockerfile to reproduce: