Update the ESS API so we can update the stored arch's should the modex include that info. Update ompi/proc to check/set the arch for remote procs, and add that function call to mpi_init right after the modex is done.
Setup to allow other grpcomm modules to decide whether or not to add the arch to the modex, and to detect if other entries have been made. If not, then the modex can just fall through. Begin setting up some logic in the "basic" module to handle different arch situations.
For now, default to the "bad" module so we will work in all situations, even though we may be sending around more info than we really require.
This fixes ticket #1340
This commit was SVN r18673.
Some minor changes to help facilitate debugger support so that both mpirun and yod can operate with it. Still to be completed.
This commit was SVN r18664.
This commit repairs the debugger initialization procedure. I am not closing the ticket, however, pending Jeff's review of how it interfaces to the ompi_debugger code he implemented. There were duplicate symbols being created in that code, but not used anywhere. I replaced them with the ORTE-created symbols instead. However, since they aren't used anywhere, I have no way of checking to ensure I didn't break something.
So the ticket can be checked by Jeff when he returns from vacation... :-)
This commit was SVN r18625.
The following Trac tickets were found above:
Ticket 1255 --> https://svn.open-mpi.org/trac/ompi/ticket/1255
After much work by Jeff and myself, and quite a lot of discussion, it has become clear that we simply cannot resolve the infinite loops caused by RML-involved subsystems calling orte_output. The original rationale for the change to orte_output has also been reduced by shifting the output of XML-formatted vs human readable messages to an alternative approach.
I have globally replaced the orte_output/ORTE_OUTPUT calls in the code base, as well as the corresponding .h file name. I have test compiled and run this on the various environments within my reach, so hopefully this will prove minimally disruptive.
This commit was SVN r18619.
Add a new function to opal_progress that tells us our recursion depth to support that solution.
Yes, I know this sounds picky, but good ol' Jeff managed to make it happen by driving his cluster near to death...
Also ensure that we declare "failed" for the daemon job when daemons fail instead of the application job. This is important so that orte knows that it cannot use xcast to tell daemons to "exit", nor should it expect all daemons to respond. Otherwise, it is possible to hang.
After lots of testing, decide to default (again) to slurm detecting failed orteds. This proved necessary to avoid rather annoying hangs that were difficult to recover from. There are conditions where slurm will fail to launch all daemons (slurm folks are working on it), and yet again, good ol' Jeff managed to find both of them.
Thanks you Jeff! :-/
This commit was SVN r18611.
Also detect orted failed-to-start by setting timeout on launch. Currently only used in TM launcher.
Neither detection is enabled by default, but are only active if heartrate is set and/or launch timeout is set. Exception for SLURM as orted failure is always detected and reported.
More info to come on devel list.
This commit was SVN r18555.
1. it depends upon the ability of the native environment to alert us that the orted has died/failed to start. I have included that support for SLURM, but other environments need to be done.
2. for some yet-to-be-determined reason, the message that tells the remaining daemons to "die" isn't getting out of the RML, even though no obvious blockage is standing in the way. Work will continue on resolving that problem. For now, the orteds appear to be exiting on their own quite nicely when they see their HNP "lifeline" disappear.
This represents the best-available fix for ticket #221 so I am closing that ticket at this time.
This commit was SVN r18536.
The former will return a valid item in the list, the latter will return an invalid item that marks the end of the list.
It was happending that when oversubscribing by way of an appfile we would cause a segv because we tried to interpret the invalid item returned by "opal_list_get_end" instead of a valid item. We would then try to write to unallocated memory.
This commit fixes trac:1279
This commit was SVN r18529.
The following Trac tickets were found above:
Ticket 1279 --> https://svn.open-mpi.org/trac/ompi/ticket/1279
such, the commit message back to the master SVN repository is fairly
long.
= ORTE Job-Level Output Messages =
Add two new interfaces that should be used for all new code throughout
the ORTE and OMPI layers (we already make the search-and-replace on
the existing ORTE / OMPI layers):
* orte_output(): (and corresponding friends ORTE_OUTPUT,
orte_output_verbose, etc.) This function sends the output directly
to the HNP for processing as part of a job-specific output
channel. It supports all the same outputs as opal_output()
(syslog, file, stdout, stderr), but for stdout/stderr, the output
is sent to the HNP for processing and output. More on this below.
* orte_show_help(): This function is a drop-in-replacement for
opal_show_help(), with two differences in functionality:
1. the rendered text help message output is sent to the HNP for
display (rather than outputting directly into the process' stderr
stream)
1. the HNP detects duplicate help messages and does not display them
(so that you don't see the same error message N times, once from
each of your N MPI processes); instead, it counts "new" instances
of the help message and displays a message every ~5 seconds when
there are new ones ("I got X new copies of the help message...")
opal_show_help and opal_output still exist, but they only output in
the current process. The intent for the new orte_* functions is that
they can apply job-level intelligence to the output. As such, we
recommend that all new ORTE and OMPI code use the new orte_*
functions, not thei opal_* functions.
=== New code ===
For ORTE and OMPI programmers, here's what you need to do differently
in new code:
* Do not include opal/util/show_help.h or opal/util/output.h.
Instead, include orte/util/output.h (this one header file has
declarations for both the orte_output() series of functions and
orte_show_help()).
* Effectively s/opal_output/orte_output/gi throughout your code.
Note that orte_output_open() takes a slightly different argument
list (as a way to pass data to the filtering stream -- see below),
so you if explicitly call opal_output_open(), you'll need to
slightly adapt to the new signature of orte_output_open().
* Literally s/opal_show_help/orte_show_help/. The function signature
is identical.
=== Notes ===
* orte_output'ing to stream 0 will do similar to what
opal_output'ing did, so leaving a hard-coded "0" as the first
argument is safe.
* For systems that do not use ORTE's RML or the HNP, the effect of
orte_output_* and orte_show_help will be identical to their opal
counterparts (the additional information passed to
orte_output_open() will be lost!). Indeed, the orte_* functions
simply become trivial wrappers to their opal_* counterparts. Note
that we have not tested this; the code is simple but it is quite
possible that we mucked something up.
= Filter Framework =
Messages sent view the new orte_* functions described above and
messages output via the IOF on the HNP will now optionally be passed
through a new "filter" framework before being output to
stdout/stderr. The "filter" OPAL MCA framework is intended to allow
preprocessing to messages before they are sent to their final
destinations. The first component that was written in the filter
framework was to create an XML stream, segregating all the messages
into different XML tags, etc. This will allow 3rd party tools to read
the stdout/stderr from the HNP and be able to know exactly what each
text message is (e.g., a help message, another OMPI infrastructure
message, stdout from the user process, stderr from the user process,
etc.).
Filtering is not active by default. Filter components must be
specifically requested, such as:
{{{
$ mpirun --mca filter xml ...
}}}
There can only be one filter component active.
= New MCA Parameters =
The new functionality described above introduces two new MCA
parameters:
* '''orte_base_help_aggregate''': Defaults to 1 (true), meaning that
help messages will be aggregated, as described above. If set to 0,
all help messages will be displayed, even if they are duplicates
(i.e., the original behavior).
* '''orte_base_show_output_recursions''': An MCA parameter to help
debug one of the known issues, described below. It is likely that
this MCA parameter will disappear before v1.3 final.
= Known Issues =
* The XML filter component is not complete. The current output from
this component is preliminary and not real XML. A bit more work
needs to be done to configure.m4 search for an appropriate XML
library/link it in/use it at run time.
* There are possible recursion loops in the orte_output() and
orte_show_help() functions -- e.g., if RML send calls orte_output()
or orte_show_help(). We have some ideas how to fix these, but
figured that it was ok to commit before feature freeze with known
issues. The code currently contains sub-optimal workarounds so
that this will not be a problem, but it would be good to actually
solve the problem rather than have hackish workarounds before v1.3 final.
This commit was SVN r18434.
* Remove the opal_only option. This was suffering from bit rot, and no one uses it. It can be added back fairly easily if wanted.
* Cleanup metadata interactions at the local level.
* Touch up some of the INC funcitonality (fix typos and a minor ordering issue)
This commit was SVN r18416.
All spawned procs must decode the port of the spawning process so they can communicate in direct routed mode.
This fixes comm_spawn for all routing modes.
This commit was SVN r18395.
Make it all work with comm_spawn in the case of all procs on previously occupied nodes, some new procs on new nodes, and mixtures of the two.
Note: comm_spawn now works with both binomial and linear routed modules. There remains a problem of spawned procs not properly getting updated contact info for the parent proc when run in the direct routed mode...but that's for another day.
This commit was SVN r18385.
If no AMCA parameters are passed then do not send across the path information. Only place it on the command line if the AMCA parameter is set.
This commit was SVN r18382.
http://www.open-mpi.org/community/lists/devel/2008/04/3779.php
{{{
svn merge -r 18276:18380 https://svn.open-mpi.org/svn/ompi/tmp-public/jjh-mca-play .
}}}
Any components not in the trunk, but in one of the effected frameworks *must* be
updated. Contact the list, look at the RFC, or look at the diff for how to do this.
Sorry for the early commit of this, but I wanted to get it in today (per RFC) and
didn't know if I would have a chance later today.
This commit was SVN r18381.
Modify grpcomm xcast so it now uses the selected routed module - eliminates cross-wiring of xcast and routing paths. Suboptimal at the moment, but better implementation is on its way.
Cleanup ignore properties on the new routed components.
This commit was SVN r18377.
Add two new API's to the routed framework - stub them out so that collaborators can work on them in various components without conflicts.
Remove a "finalize" from the select function that could cause problems as the component had not had its initialize called yet.
This commit was SVN r18369.
Update the rsh tree spawn capability so we spawn the next wave of daemons before launching our own local procs.
Add an ability to encode nodenames for large clusters with contiguous node name numbering schemes - this allows communication of all node names in a few bytes instead of tens-of-bytes/node.
This commit was SVN r18338.
The problem was caused by a bad ordering between the restart of the ORTE level tcp connections (in the OOB - out-of-band communication) and the Open MPI level tcp connections (BTLs). Before this commit ORTE would shutdown and restart the OOB completely before the OMPI level restarted its tcp connections. What would happen is that a socket descriptor used by the OMPI level on checkpoint was assigned to the ORTE level on restart. But the OMPI level had no knowledge that the socket descriptor it was previously using has been recycled so it closed it on restart. This caused the ORTE level to break as the newly created socket descriptor was closed without its knowledge.
The fix is to have the OMPI level shutdown tcp connections, allow the ORTE level to restart, and then allow the OMPi level to restart its connections. This seems obvious, and I'm surprised that this bug has not cropped up sooner. I'm confident that this specific problem has been fixed with this commit.
Thanks to Eric Roman and Tamer El Sayed for their help in identifying this problem, and patience while I was fixing it.
* Add a new state {{{OPAL_CRS_RESTART_PRE}}}. This state identifies when we are on the down slope of the INC (finalize-like) which is useful when you want to close, but not reopen a component set for fear of interfering with a lower level.
* Use this new state in OMPI level coordination. Here we want to make sure to play well with both the OMPI/BTL/TCP and ORTE/OOB/TCP components.
* Update ft_event functions in PML and BML to handle the new restart state.
* Add an additional flag to the error output in OOB/TCP so we can see what the socket descriptor was on failure as this can be helpful in debugging.
This commit was SVN r18276.
We were mistakenly computing the local_rank across -all- jobs with procs on that node. While the two definitions are equivalent for an initial launch, comm_spawn'd procs would get the wrong local_rank. In particular, there would not be a local_rank=0 proc in the comm_spawn'd job on any node that was shared with the initial job.
This commit was SVN r18263.
Fix a potential problem with RM-provided nodenames not matching returns from gethostname - ensure that the HNP's nodename gets DNS-resolved when comparing against RM-provided hostnames. Note that this may be an issue for RM-based clusters that don't have local DNS resolution, but hopefully that is more indicative of a poorly configured system.
This commit was SVN r18252.
{{{
svn merge -r 18218:18240 https://svn.open-mpi.org/svn/ompi/tmp/jjh-scratch .
}}}
Contains:
* Primarily a fix for a user reported problem where a cached file descriptor is causing a SIGPIPE on restart.
* Cleanup some small memory leaks from using mca_base_param_env_var() - Thanks Jeff
* Cleanup ORTE FT tool compilation in non-FT builds - Thanks Tim P.
* Cleanup mpi interface with missplaced {{{OPAL_CR_ENTER_LIBRARY}}} - Thanks Terry
* Some other sundry cleanup items all dealing with C/R functionality in the trunk.
This commit was SVN r18241.
Restore the "do-not-launch" functionality so users can test a mapping without launching it.
Add a "do-not-resolve" cmd line flag to mpirun so the opal/util/if.c code does not attempt to resolve network addresses, thus enabling a user to test a hostfile mapping without hanging on network resolve requests.
Add a function to hostfile to generate an ordered list of host names from a hostfile
This commit was SVN r18190.
Fix the ompi-server -h cmd line option so it actually tells you something!
Add two new testing codes to the orte/test/mpi area: accept and connect.
This commit was SVN r18176.
Like btl_tcp_disable_family, this parameter more or less disables
a whole address family. Though the sockets are still created, the
corresponding information isn't added to the connection strings.
Likewise, we don't try to connect to addresses matching the disabled
address family.
This is particularly important for multidomain clusters, where IPv4 is
oftenly filtered (firewalled), sometimes by simply dropping the packets
instead of rejecting them (thus causing a connection timeout instead of
a quick "no route to host").
This commit was SVN r18163.
Add the daemon map capability to the ODLS to create and save a map of daemon vpid vs nodename from the launch message.
Cleanup a few places in the base plm launch support where we didn't adequately protect rml recv's from potentially executing sends.
This commit was SVN r18143.
1. applied prefix rule to functions and variables of RMAPS rank_file component
2. cleaned ompi_mpi_init.c from paffinity code
3. paffinity code moved to new opal/mca/paffinity/base/paffinity_base_service.c file
4. added opal_paffinity_slot_list mca parameter
This commit was SVN r18019.
The bug was a race condition in the barrier operation that caused the barrier in MPI_Finalize to fail on very short programs.
Scalaiblity was improved by using the daemons to aggregate modex and barrier messages before sending them to the rank=0 proc. Improvement is proportional to ppn, of course, but there really wasn't a scaling problem at low ppn anyway. This modification also paves the way for better allgather operations since now all the data for each node is sitting at the daemon level, and the daemons are now aware that a collective operation on the OOB is underway (so they -can- participate in a collective of their own to support it).
Also added better diagnostics to map out the timing associated with MPI_Init - turned on by -mca orte_timing 1.
This commit was SVN r17988.
Event uninit_use: Using uninitialized value "rc"
Instead of initializing rc in the beginning, rather use return value
of opal_hash_table_set_value_uint32.
This commit was SVN r17976.
Specifically, add two new APIs:
1. lost_route: allows the OOB to report that a connection has failed, thereby giving the routed module an opportunity to respond appropriately to its topology. Creating the API also allows each routed component to hold its own definition of "lifeline" - in some cases, this may be a single connection, but in others it may be multiple connections. Some modules may choose to re-route messaging if the lifeline or any other connection is lost, while others may choose to abort the job.
Both the tree and unity modules retain the current behavior and abort the job if the lifeline connection is lost, while ignoring other lost connections.
2. get_wireup_info: returns (in a provided buffer) info required to wireup connections for the specified job. Some routed modules do not need to return any info as they can wireup via alternative means, while some need to xchg data with their peers. If info is inserted into the buffer, the plm_base_launch_apps function will xcast the contents to the specified job.
The commit also removes the "lifeline" entry from the orte_process_info struct (and the associated ORTE_PROC_MY_LIFELINE definition) as the lifeline info is now contained within the respective routed module.
This commit was SVN r17969.
Clarify the setting of send_first in the mpi bindings (trivial, i know, but helpful)
Remove the extra xcast of child contact info to the parent job.
This commit was SVN r17952.
Reogranize the grpcomm code a little to provide support for soon-to-come new grpcomm components. The revised organization puts what will be common code elements in the base to avoid duplication, while allowing components that don't need those functions to ignore them.
This commit was SVN r17941.
Only one place used the user name field - session_dir, when formulating the name of the top-level directory. Accordingly, the code for getting the user's id has been moved to the session_dir code.
This commit was SVN r17926.
This has been a long-time problem. I tried to reduce the problem by having the orteds tell the HNP they were finalizing, and having the HNP wait until all orteds had reported or we timed out.
What was observed was that all the orteds were correctly reporting that they are leaving, but the HNP is able to exit before the orteds, thus closing the orteds lifeline socket and generating the error output. This is caused by the fact that the orteds have to whack all remaining session directories, which includes that blasted monster shared memory file! Cleaning up the SM file can take quite a while.
The HNP doesn't have that problem as there is no SM file there! So it gets out first.
What we had done in the past to resolve that problem was put a little test in the OOB that checks to see if we are finalizing. If we are, then we ignore the lifeline connection being lost. That check was still in the code - however, we had lost the line in orte_finalize that set the flag!!
This commit was SVN r17893.
Fix race conditions in abnormal terminations. We had done a first-cut at this in a prior commit. However, the window remained partially open due to the fact that the HNP has multiple paths leading to orte_finalize. Most of our frameworks don't care if they are finalized more than once, but one of them does, which meant we segfaulted if orte_finalize got called more than once. Besides, we really shouldn't be doing that anyway.
So we now introduce a set of atomic locks that prevent us from multiply calling abort, attempting to call orte_finalize, etc. My initial tests indicate this is working cleanly, but since it is a race condition issue, more testing will have to be done before we know for sure that this problem has been licked.
Also, some updates relevant to the tool comm library snuck in here. Since those also touched the orted code (as did the prior changes), I didn't want to attempt to separate them out - besides, they are coming in soon anyway. More on them later as that functionality approaches completion.
This commit was SVN r17843.
Comm_spawn was sticking during spawn_multiple because of a problem in the dpm - the modex there is asking processes to talk to each other in an allgather_list operation, but the procs don't have the required contact info to do so. The solution here was to ensure that all parent procs have full contact info for procs in the child job.
Admittedly, this isn't the long-term answer. We would like to have the contact info given to only the parent procs that were involved in the comm_spawn. There is a way to do that, but this will suffice to keep things working until that can be implemented and tested.
This commit was SVN r17772.
Ensure that direct xcast handles all its use-cases correctly.
Unity routed component needs to use the base recv function to properly operate.
This commit was SVN r17764.
This commit adds definition for a "lifeline" connection. For an HNP, there is no lifeline, so the lifeline proc is NULL. For a daemon, the lifeline is the HNP - the daemon should abort if it loses that connection.
For a proc using unity routed, the lifeline is the HNP since it connects directly to the HNP.
For a proc using tree routed, the lifeline is the local daemon.
Adjusted OOB to call abort if the lifeline (as opposed to HNP) connection is lost.
This commit was SVN r17761.
The change also:
- cleans up and simplifies the command line processing code
- adds an error output if more than one hostfile passed for a single app context
- gets rid of the superfluous orte_app_context_map_t type, and instead use a simple argv of -host options
This commit was SVN r17750.
The following Trac tickets were found above:
Ticket 1124 --> https://svn.open-mpi.org/trac/ompi/ticket/1124
Note that --path specifies extra directories where the executable
is searched for, but does not affect the PATH settings.
This commit fixes trac:1221.
This commit was SVN r17748.
The following Trac tickets were found above:
Ticket 1221 --> https://svn.open-mpi.org/trac/ompi/ticket/1221
* Extension to the ESS framework to support C/R
* Fixed support for {{{snapc_base_establish_global_snapshot_dir}}}
* Fixed FileM support
* Misc. minor code modifications
There are some outstanding visability issues that I want to fix next.
This commit was SVN r17725.
Also, update some properties (source files should not be executeable...), and remove a couple unneeded inclusions of orte_proc_table.h
This commit was SVN r17655.
Basically, the method employed here is to have a recv create a zero-time timer event that causes the event library to execute a function that processes the message once the recv returns. Thus, any action taken as a result of processing the message occur outside of a recv.
Created two new macros to assist:
ORTE_MESSAGE_EVENT: creates the zero-time event, passing info in a new orte_message_event_t object
ORTE_PROGRESSED_WAIT: while waiting for specified conditions, just calls progress so messages can be recv'd.
Also fixed the failed_launch function as we no longer block in the orted callback function. Updated the error messages to reflect revision. No change in API to this function, but PLM "owners" may want to check their internal error messages to avoid duplication and excessive output.
This has been tested on Mac, TM, and SLURM.
This commit was SVN r17647.
about linkers, have all OPAL, ORTE, and OMPI components '''not'' link
against the OPAL, ORTE, or OMPI libraries.
See ttp://www.open-mpi.org/community/lists/users/2007/10/4220.php for
details (or https://svn.open-mpi.org/trac/ompi/wiki/Linkers for a
better-formatted version of the same info).
This commit was SVN r16968.
This commit also cleans up the checkpoint and terminate case making it more
precise than before. Previously the application could make a small amount of
progress between checkpoint completion and application termination. Now the
application will make no progress at all in this time span.
Additional minor change:
- Start using OPAL_INT_TO_BOOL instead of if/else logic
This commit was SVN r16952.
has his own range which is defined by a min value and a range. By default
there is no limitation on the port range, which is exactly the same
behavior as before.
This commit was SVN r16584.
If we cannot resolve the route to the peer that we're trying to send
to, don't queue up the message in the TCP OOB -- instead, return it to
the upper layer (e.g., the RML) and let it decide what to do.
In the case of the routed RML, the tree component will queue it up for
later transmission. Hence, we don't want the message queued up both
here in the TCP OOB and the tree routed. Also see some more
discussion / explanation in #1171.
This commit was SVN r16540.
The following SVN revision numbers were found above:
r16513 --> open-mpi/ompi@7ae9589d70
The following Trac tickets were found above:
Ticket 1170 --> https://svn.open-mpi.org/trac/ompi/ticket/1170
Note that this means ALL procs in the parent job are updated, even though they may not be participating in the comm_spawn. This doesn't really hurt anything - just unnecessary.
Comm_spawn still has a problem when a child process shares a node with a parent, so this doesn't fix everything. It only fixes the bug of ensuring all procs know how to talk to each other.
This commit was SVN r16460.
This commit introduces the necessary logic to avoid that conflict. If a PLS component can identify that a daemon has failed, then we will set a flag indicating that fact. The xcast system will subsequently check that flag and, if it is set, will send all messages direct to the recipient. In the case of "kill local procs" and "terminate", the messages will go directly to each orted, thus bypassing any orted that has failed.
In addition, the xcast system will -not- wait for the messages to complete, but will return immediately (i.e., operate in non-blocking mode). Orterun will wait (via an event timer) for a period of time based on the number of daemons in the system to allow the messages to attempt to be delivered - at the end of that time, orterun will simply exit, alerting the user to the problem and -strongly- recommending they run orte-clean.
I could only test this on slurm for the case where all daemons unexpectedly died - srun apparently only executes its waitpid callback when all launched functions terminate. I have asked that Jeff integrate this capability into the OOB as he is working on it so that we execute it whenever a socket to an orted is unexpectedly closed. Meantime, the functionality will rarely get called, but at least the logic is available for anyone whose environment can support it.
This commit was SVN r16451.
variable is not defined. Make sure to set it to something reasonable
so that file preloading still works (instead of seg faulting :)
Thanks to Hiep Bui Hoang for reporting this bug.
This commit was SVN r16433.
1. taking advantage of the fact that we no longer create the launch message via a GPR trigger. In earlier times, we had the GPR create the launch message based on a subscription. In that mode of operation, we could not guarantee the order in which the data was stored in the message - hence, we had no choice but to parse the message in a loop that checked each value against a list of possible "keys" until the corresponding value was found.
Now, however, we construct the message "by hand", so we know precisely what data is in each location in the message. Thus, we no longer need to send the character string "keys" for each data value any more. This represents a rather large savings in the message size - to give you an example, we typically would use a 30-char "key" for a 2-byte data value. As you can see, the overhead can become very large.
2. sending node-specific data only once. Again, because we used to construct the message via subscriptions that were done on a per-proc basis, the data for each node (e.g., the daemon's name, whether or not the node was oversubscribed) would be included in the data for each proc. Thus, the node-specific data was repeated for every proc.
Now that we construct the message "by hand", there is no reason to do this any more. Instead, we can insert the data for a specific node only once, and then provide the per-proc data for that node. We therefore not only save all that extra data in the message, but we also only need to parse the per-node data once.
The savings become significant at scale. Here is a comparison between the revised trunk and the trunk prior to this commit (all data was taken on odin, using openib, 64 nodes, unity message routing, tested with application consisting of mpi_init/mpi_barrier/mpi_finalize, all execution times given in seconds, all launch message sizes in bytes):
Per-node scaling, taken at 1ppn:
#nodes original trunk revised trunk
time size time size
1 0.10 819 0.09 564
2 0.14 1070 0.14 677
3 0.15 1321 0.14 790
4 0.15 1572 0.15 903
8 0.17 2576 0.20 1355
16 0.25 4584 0.21 2259
32 0.28 8600 0.27 4067
64 0.50 16632 0.39 7683
Per-proc scaling, taken at 64 nodes
ppn original trunk revised trunk
time size time size
1 0.50 16669 0.40 7720
2 0.55 32733 0.54 11048
3 0.87 48797 0.81 14376
4 1.0 64861 0.85 17704
Condensing those numbers, it appears we gained:
per-node message size: 251 bytes/node -> 113 bytes/node
per-proc message size: 251 bytes/proc -> 52 bytes/proc
per-job message size: 568 bytes/job -> 399 bytes/job
(job-specific data such as jobid, override oversubscribe flag, total #procs in job, total slots allocated)
The fact that the two pre-commit trunk numbers are the same confirms the fact that each proc was containing the node data as well. It isn't quite the 10x message reduction I had hoped to get, but it is significant and gives much better scaling.
Note that the timing info was, as usual, pretty chaotic - the numbers cited here were typical across several runs taken after the initial one to avoid NFS file positioning influences.
Also note that this commit removes the orte_process_info.vpid_start field and the handful of places that passed that useless value. By definition, all jobs start at vpid=0, so all we were doing is passing "0" around. In fact, many places simply hardwired it to "0" anyway rather than deal with it.
This commit was SVN r16428.
1. --with-sge, always builds
2. --without-sge, never builds
3. if neither is specified, build if and only if either SGE_ROOT is set or "qrsh" is found in the path
This commit was SVN r16422.
This patch also fixes a minor bug discovered along the way: we had "lost" the passing of the oversubscribed condition flag from the mapper to the orteds. Thus, we were not setting sched_yield correctly when in oversubscribed conditions (except when a hostfile was specified - different logic there because we treat the number of slots allocated on the node as "uncertain")
I did not modify the process component in this patch - I will send a proposed patch to the maintainers of that component so they can review it first.
This commit was SVN r16418.
* Fix some missing includes in a few places.
* Add the cr_request() functionality to the BLCR CRS component.
We are now dependent upon the 0.6.* series of BLCR.
* Made the CR notification mechanism a registered function.
This way we can have an OPAL-only version and it can be replaced at
runtime with the ORTE version.
* Add a 'opal_cr_allow_opal_only' parameter that will enable OPAL-only
CR functionality when the user wants it. Default: Disabled.
* Fix the placement of a checkpoint request check in MPI_Init
* Pull the OPAL notification mechanism into the SnapC framework.
* We no longer fork/exec the 'opal-checkpoint' command for local
checkpointing, the Local coordinator in the orted does this directly.
* The Local and Application coordinator talk together bypassing the OPAL
notifiation mechanism.
* Optimized the Local <-> App Coordinator communication.
* Improved the structure used to track vpid_snapshots in the local coord.
* Fix a race condition in which an application under heavy communication load
may produce an inconsistent global checkpoint.
This commit was SVN r16389.
The commit has been tested for C/R and Cray operations, and on Odin (SLURM, rsh) and RoadRunner (TM). I tried to update all environments, but obviously could not test them. I know that Windows needs some work, and have highlighted what is know to be needed in the odls process component.
This represents a lot of work by Brian, Tim P, Josh, and myself, with much advice from Jeff and others. For posterity, I have appended a copy of the email describing the work that was done:
As we have repeatedly noted, the modex operation in MPI_Init is the single greatest consumer of time during startup. To-date, we have executed that operation as an ORTE stage gate that held the process until a startup message containing all required modex (and OOB contact info - see #3 below) info could be sent to it. Each process would send its data to the HNP's registry, which assembled and sent the message when all processes had reported in.
In addition, ORTE had taken responsibility for monitoring process status as it progressed through a series of "stage gates". The process reported its status at each gate, and ORTE would then send a "release" message once all procs had reported in.
The incoming changes revamp these procedures in three ways:
1. eliminating the ORTE stage gate system and cleanly delineating responsibility between the OMPI and ORTE layers for MPI init/finalize. The modex stage gate (STG1) has been replaced by a collective operation in the modex itself that performs an allgather on the required modex info. The allgather is implemented using the orte_grpcomm framework since the BTL's are not active at that point. At the moment, the grpcomm framework only has a "basic" component analogous to OMPI's "basic" coll framework - I would recommend that the MPI team create additional, more advanced components to improve performance of this step.
The other stage gates have been replaced by orte_grpcomm barrier functions. We tried to use MPI barriers instead (since the BTL's are active at that point), but - as we discussed on the telecon - these are not currently true barriers so the job would hang when we fell through while messages were still in process. Note that the grpcomm barrier doesn't actually resolve that problem, but Brian has pointed out that we are unlikely to ever see it violated. Again, you might want to spend a little time on an advanced barrier algorithm as the one in "basic" is very simplistic.
Summarizing this change: ORTE no longer tracks process state nor has direct responsibility for synchronizing jobs. This is now done via collective operations within the MPI layer, albeit using ORTE collective communication services. I -strongly- urge the MPI team to implement advanced collective algorithms to improve the performance of this critical procedure.
2. reducing the volume of data exchanged during modex. Data in the modex consisted of the process name, the name of the node where that process is located (expressed as a string), plus a string representation of all contact info. The nodename was required in order for the modex to determine if the process was local or not - in addition, some people like to have it to print pretty error messages when a connection failed.
The size of this data has been reduced in three ways:
(a) reducing the size of the process name itself. The process name consisted of two 32-bit fields for the jobid and vpid. This is far larger than any current system, or system likely to exist in the near future, can support. Accordingly, the default size of these fields has been reduced to 16-bits, which means you can have 32k procs in each of 32k jobs. Since the daemons must have a vpid, and we require one daemon/node, this also restricts the default configuration to 32k nodes.
To support any future "mega-clusters", a configuration option --enable-jumbo-apps has been added. This option increases the jobid and vpid field sizes to 32-bits. Someday, if necessary, someone can add yet another option to increase them to 64-bits, I suppose.
(b) replacing the string nodename with an integer nodeid. Since we have one daemon/node, the nodeid corresponds to the local daemon's vpid. This replaces an often lengthy string with only 2 (or at most 4) bytes, a substantial reduction.
(c) when the mca param requesting that nodenames be sent to support pretty error messages, a second mca param is now used to request FQDN - otherwise, the domain name is stripped (by default) from the message to save space. If someone wants to combine those into a single param somehow (perhaps with an argument?), they are welcome to do so - I didn't want to alter what people are already using.
While these may seem like small savings, they actually amount to a significant impact when aggregated across the entire modex operation. Since every proc must receive the modex data regardless of the collective used to send it, just reducing the size of the process name removes nearly 400MBytes of communication from a 32k proc job (admittedly, much of this comm may occur in parallel). So it does add up pretty quickly.
3. routing RML messages to reduce connections. The default messaging system remains point-to-point - i.e., each proc opens a socket to every proc it communicates with and sends its messages directly. A new option uses the orteds as routers - i.e., each proc only opens a single socket to its local orted. All messages are sent from the proc to the orted, which forwards the message to the orted on the node where the intended recipient proc is located - that orted then forwards the message to its local proc (the recipient). This greatly reduces the connection storm we have encountered during startup.
It also has the benefit of removing the sharing of every proc's OOB contact with every other proc. The orted routing tables are populated during launch since every orted gets a map of where every proc is being placed. Each proc, therefore, only needs to know the contact info for its local daemon, which is passed in via the environment when the proc is fork/exec'd by the daemon. This alone removes ~50 bytes/process of communication that was in the current STG1 startup message - so for our 32k proc job, this saves us roughly 32k*50 = 1.6MBytes sent to 32k procs = 51GBytes of messaging.
Note that you can use the new routing method by specifying -mca routed tree - if you so desire. This mode will become the default at some point in the future.
There are a few minor additional changes in the commit that I'll just note in passing:
* propagation of command line mca params to the orteds - fixes ticket #1073. See note there for details.
* requiring of "finalize" prior to "exit" for MPI procs - fixes ticket #1144. See note there for details.
* cleanup of some stale header files
This commit was SVN r16364.
