1
1

472 Коммитов

Автор SHA1 Сообщение Дата
Ralph Castain
92c996487c Update how we pass the node regex so we pass _all_ nodes, even those without daemons. This allows the backend daemons to form a complete picture of the allocation. Include info on which nodes have daemons on them, and populate that info on the backend as well.
Set the daemons' state to "running" and mark them as "alive" by default when constructing the nidmap

Get the DVM running again

Fix direct modex by eliminating race condition caused by releasing data while sending it

Up the size limit before compressing

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-04-03 19:25:15 -07:00
Ralph Castain
2cc5fea8be Update to PMIx v2.0alpha
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-04-03 10:02:29 -07:00
Ralph Castain
7dd34d0c9a Use the correct callback data - the callback function was expecting a bool*, not a pmix_ptl_sr_t*.
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-03-28 17:21:47 -07:00
Ralph Castain
35f817911e Fix coverity issues
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-03-24 08:09:46 -07:00
Ralph Castain
c0bcd11bcf Fix permissions - no CI required
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-03-23 08:05:52 -07:00
Ralph Castain
55e4fba5f5 If we lose connection to the server after initiating a send/recv in PMIx (e.g., in PMIx_Abort), then we need to "resolve" all pending recvs to avoid hanging.
Fixes #3225

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-03-23 02:53:21 -07:00
Ralph Castain
d645557fa0 Update to include the PMIx 2.0 APIs for monitoring and job control. Include required integration, but leave the monitors off for now. Move the sensor framework out of ORTE as it is being absorbed into PMIx
Fix typo and silence warnings

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-03-21 17:47:08 -07:00
Ralph Castain
4b6d220a83 You cannot include both pmi.h and pmi2.h as they have conflicting defines in them.
Thanks to Kilian Cavalotti for pointing it out

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-03-19 11:53:54 -07:00
Ralph Castain
c6bc3ccb76 Sync to latest PMIx master and PMIx reference server
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-03-11 12:50:38 -08:00
Ralph Castain
aca7091114 Fix some minor compatibility issues by ensuring job-level data gets stored against wildcard rank in the cray, s1, and s2 components, and that the ext1 component translates all wildcard rank requests into the peer's rank since v1.x of PMIx doesn't understand wildcard ranks
Closes #3101

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-03-05 10:30:59 -08:00
Ralph Castain
1de72ff023 Silence an unnecessary error log
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-03-02 17:18:34 -08:00
Ralph Castain
e86a0dbf39 Update to PMIx master to include dlopen fixes and addition of libltdl support
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-02-22 11:54:33 -08:00
Ralph Castain
8cffdcf127 Ensure that the pmix headers and lib get installed when --with-devel-headers is given so that PMIx applications can be built and executed against the "embedded" PMIx version
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-02-21 13:46:46 -08:00
Gilles Gouaillardet
bb2481a84b pmix2x: synchronize to the latest PMIx master
pmix/master@f57d9b2953

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-02-20 10:45:17 +09:00
Ralph Castain
f49118eaab Fix some pmix configuration code
Remove stale file reference that caused a check to always fail. Update psm2 function check to new libs

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-02-16 10:54:47 -08:00
Howard Pritchard
b272f87926 Merge pull request #2968 from hjelmn/pmix_cray
pmix/cray: performance improvements and cleanup
2017-02-16 11:41:59 -07:00
Ralph Castain
201f8571ca Ensure we retain the peer object until we are done with it, then detect that the socket has closed due to a lost connection and cleanly release the message event
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-02-15 18:30:55 -08:00
Ralph Castain
9cd7349d7c Instead of completely free'ing the event base, pause the PMIx progress thread before tearing down the infrastructure, and then release the event base at the end of the procedure. This allows any infrastructure objects holding events to delete them prior to free'ing the event base.
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-02-15 05:02:43 -08:00
Ralph Castain
f7fe2f7189 Merge pull request #2977 from rhc54/topic/spawn
Fix comm_spawn by registering nspace info only when needed
2017-02-15 04:31:54 -08:00
Ralph Castain
68b53e2179 Fix comm_spawn by registering nspace info only when needed - either when we have local procs, or when job-level info is required by connecting jobs
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-02-14 19:47:56 -08:00
Ralph Castain
0c8609ca16 Update to newest PMIx master (includes configuration cleanups). Silence trivial Coverity warning in hwloc base.
Cleanup a race condition segfault during finalize by ensuring the PMIx progress thread is stopped prior to starting to tear down the messaging components

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-02-14 15:14:00 -08:00
Nathan Hjelm
3b912ea2a7 pmix/cray: performance improvements and cleanup
Do not use opal_output_verbose inside O(n) loops. This was causing us
to make O(n) calls to snprintf which was greatly slowing launch at
scale.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2017-02-14 11:13:10 -07:00
Ralph Castain
35578b4009 Update to lastest PMIx master
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-02-13 23:19:26 -08:00
Gilles Gouaillardet
7acef4833e pmix2x: Update to latest PMIx master
pmix/master@6ed27be839

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-02-08 13:23:27 +09:00
KAWASHIMA Takahiro
4b2eba34a6 Merge pull request #2933 from kawashima-fj/pr/dstore-config-desc
pmix/pmix2x: Correct configure option description
2017-02-08 13:03:27 +09:00
Jeff Squyres
100b112d3c pmix: fix zlib protection macro usage
It's possible that we can have zlib.h but still not have zlib support.
Use the correct macro to protect the usage of calling zlib functions.

This fixes 32-bit MTT builds at Cisco (e.g.,
https://mtt.open-mpi.org/index.php?do_redir=2389).

Submitted upstream to PMIX: https://github.com/pmix/master/pull/290

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-02-07 05:52:32 -08:00
KAWASHIMA Takahiro
750406f67b pmix/pmix2x: Correct configure option description
`--enable-pmix-dstore` option was enabled by default in f4a5511.

Signed-off-by: KAWASHIMA Takahiro <t-kawashima@jp.fujitsu.com>
2017-02-07 11:52:56 +09:00
Ralph Castain
edcfdf2365 Update to latest PMIx master
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-31 08:01:37 -08:00
Gilles Gouaillardet
b078e57e73 pmix/ext1x: fix misc memory leaks in namespace registration
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-30 10:52:42 +09:00
Gilles Gouaillardet
f51fc293a2 ext1x/pmix1x_client: plug misc memory leaks
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-30 10:52:42 +09:00
Gilles Gouaillardet
022cca79ea pmix/ext1x: plug a memory leak in opal_lkupcbfunc()
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-30 10:52:36 +09:00
Gilles Gouaillardet
f485d12a82 pmix: rename the ext11 component into ext1x
also use the same naming scheme thann pmix/ext2x

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-30 10:52:35 +09:00
Gilles Gouaillardet
dccb1899e6 pmix/ext11: correctly use PMIx_server_register_nspace()
PMIx_server_register_nspace() is an asynchronous operation, so
the pmix glue wait for it completes before returning.

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-30 09:23:19 +09:00
Gilles Gouaillardet
6955e1e25c pmix/ext11: fix compilation
the argc field from the opal_pmix_app_t struct was removed,
so adjust the pmix/ext11 glue accordingly.

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-30 09:23:18 +09:00
Ralph Castain
3302864a7d Cleanup a typo that can cause a segfault - use a local variable name different than the one passed into the function
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-27 16:49:25 -08:00
Gilles Gouaillardet
896434b1bd pmix/ext2x: plug a memory leak in opal_lkupcbfunc()
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-26 14:07:15 +09:00
Gilles Gouaillardet
6b8e1c217c pmix/ext2x: plug misc memory leaks
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-26 14:06:58 +09:00
Gilles Gouaillardet
142b95df87 pmix/ext2x: plug misc memory leaks regarding opal_pmix2x_event_chain_t handling
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-25 16:17:10 +09:00
Gilles Gouaillardet
7a3d39f079 pmix/ext2x: plug a memory leak in _reg_nspace()
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-25 16:17:01 +09:00
Gilles Gouaillardet
189da7fdab pmix2x: plug a memory leak in _event_hdlr()
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-24 09:13:30 +09:00
Gilles Gouaillardet
acbc32d3b2 pmix2x: plug a memory leak in opal_lkupcbfunc()
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-24 09:13:29 +09:00
Gilles Gouaillardet
b5b21043c4 pmix2x: plug a memory leak in _reg_nspace()
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-24 09:13:29 +09:00
Gilles Gouaillardet
0f47310a75 pmix2x/pmix2x_client: plug misc memory leaks
Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-24 09:13:29 +09:00
Ralph Castain
8c960bae8d Update to latest PMIx master
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-23 07:07:40 -08:00
Ralph Castain
e568b211e4 Silence Coverity CID 1398541
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-10 15:30:50 -08:00
Gilles Gouaillardet
6d59b476de Merge pull request #2686 from ggouaillardet/topic/pmix2x_ptl_base_sendrecv
pmix2x: ptl/base: send header and message data together via writev()
2017-01-10 16:26:10 +09:00
Gilles Gouaillardet
44c1ff60f1 Merge pull request #2672 from ggouaillardet/topic/misc_memory_leaks
Plug misc memory leaks
2017-01-10 13:16:04 +09:00
Gilles Gouaillardet
a01960bee5 pmix2x: ptl/base: send header and message data together via writev()
on Linux, sending the header and then the message data does severely
impact performances of ptl/tcp :
on the receiver, reading the data can often result in an PMIX_ERR_RESOURCE_BUSY
or PMIX_ERR_WOULD_BLOCK, which ends up degrading performances)
this commit send both header and message data at the same time via writev()
and makes ptl/tcp virtually as efficient as ptl/usock.

Short writev generally occur when the kernel buffer is full, so there is no
point for retrying in this case.

fwiw, no such degradation was observed on OSX.

Refs open-mpi/ompi#2657

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-10 13:07:39 +09:00
Ralph Castain
67fce2861b Merge pull request #2685 from rhc54/topic/cov
Resolve Coverity issues
2017-01-07 13:11:40 -08:00
Ralph Castain
e25e69dc2f Resolve Coverity issues
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-07 10:45:52 -08:00