Bring the ofi/rml component online #3836
Conversation
|
Does this fix the orte_no_op over ofi ? |
|
Yes - I've been launching MPI jobs with it just fine. One of the builds in Jenkins is failing, though, so I'm trying to track that down |
|
hmmm...I can't make it fail. I wonder if this is something peculiar to suse? @bwbarrett Is there any way to get more debug info out of that build? I've tried building with the same config, forcing various components off, etc. - to no avail. However, I don't have a suse build. |
…for the daemons. Cleanup the current confusion over how connection info gets created and passed to make it all flow thru the opal/pmix "put/get" operations. Update the PMIx code to latest master to pickup some required behaviors. Remove the no-longer-required get_contact_info and set_contact_info from the RML layer. Add an MCA param to allow the ofi/rml component to route messages if desired. This is mainly for experimentation at this point as we aren't sure if routing wi ll be beneficial at large scales. Leave it "off" by default. Signed-off-by: Ralph Castain <rhc@open-mpi.org>
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
| PMIX_DESTRUCT(&pbkt); | ||
| return rc; | ||
| } | ||
| <<<<<<< c632784ca34c467055eadcb4efe84a25e5a3911b:opal/mca/pmix/pmix2x/pmix/src/dstore/pmix_esh.c |
| } | ||
| } | ||
|
|
||
| <<<<<<< c632784ca34c467055eadcb4efe84a25e5a3911b:opal/mca/pmix/pmix2x/pmix/src/dstore/pmix_esh.c |
There was a problem hiding this comment.
nah, looks like i just missed removing the last line of the conflict flag - i can clean that up
There was a problem hiding this comment.
I saw multiple of them and there was clearly an unresolved merges.
There was a problem hiding this comment.
not sure that last part is true, but no matter - it runs just fine and passed an MTT scan, so I think things resolved correctly. Anyway, I'm updating from tarball now.
There was a problem hiding this comment.
https://github.com/rhc54/ompi/blob/f7e8780a42eb512ed5c54715471168814ef63383/opal/mca/pmix/pmix2x/pmix/src/mca/gds/ds12/gds_dstore.c#L2811:L2829
I'm 100% that this will not compile!
There was a problem hiding this comment.
relax dude - do you not see the .pmix_ignore in that component? It is prevented from compiling due to the reported problems in the dstore. I am removing the ignore now as it seems functional, albeit the memory footprint problem remains
There was a problem hiding this comment.
What memory footprint problem? Is there any issue on that?
I think that @karasevb have fixed (at least) one of the possible causes he discovered when keys was sent in the message from the server to the client even if the dstore was used.
There was a problem hiding this comment.
Look at the issue he filed this morning in PMIx
There was a problem hiding this comment.
Ok, thanks. I thought it's something else.
3 participants
Still only used if rml_ofi_desired=1