5.3. User-Level Fault Mitigation (ULFM)

This chapter documents the features and options specific to the User Level Failure Mitigation (ULFM) Open MPI implementation.

5.3.1. Quick Start

This is an extremely terse summary of how to use ULFM:

shell$ ./configure --with-ft=ulfm [...options...]
shell$ make [-j N] all install
shell$ mpicc my-ft-program.c -o my-ft-program
shell$ mpirun -n 4 --with-ft ulfm my-ft-program

5.3.2. Features

This implementation conforms to the User Level Failure Mitigation (ULFM) MPI Standard draft proposal. The ULFM proposal is developed by the MPI Forum’s Fault Tolerance Working Group to support the continued operation of MPI programs after any type of failures, hard or soft, have impacted the execution. The key principle is that no MPI call (point-to-point, collective, RMA, IO, …) can block indefinitely after a failure, but must either succeed or raise an MPI error. Accordingly, the errors are not all fatal, the MPI implementations will do a best-effort approach to maintain the execution environment up and running.

This implementation produces the three supplementary error codes and five supplementary interfaces defined in the communicator section of the ULFM chapter standard draft document.

  • MPIX_ERR_PROC_FAILED when a process failure prevents the completion of an MPI operation (error code).

  • MPIX_ERR_PROC_FAILED_PENDING when a potential sender matching a non-blocking wildcard source receive has failed (error code).

  • MPIX_ERR_REVOKED when the communicator is revoked (error code).

  • MPIX_Comm_revoke(MPI_Comm comm) Interrupts any communication pending on the communicator at all ranks (API). See MPIX_Comm_revoke.

  • MPIX_Comm_is_revoked(MPI_Comm comm, int *flag) Test if a Communicator is currently revoked (API). See MPIX_Comm_is_revoked.

  • MPIX_Comm_shrink(MPI_Comm comm, MPI_Comm* newcomm) creates a new communicator where dead processes in comm were removed, and the remaining processes are renamed to cover all the gaps in the naming from the original communicator (API). See MPIX_Comm_shrink, MPIX_Comm_ishrink.

  • MPIX_Comm_agree(MPI_Comm comm, int *flag) performs a consensus (i.e. fault tolerant allreduce operation) on flag (with the operation bitwise AND) (API). Absorbs all new failures, and propagate the knowledge about failures among the participants. see MPIX_Comm_agree, MPIX_Comm_iagree.

  • MPIX_Comm_get_failed(MPI_Comm comm, MPI_Group* failedgrp) obtains the group of currently failed processes (API). See MPIX_Comm_get_failed.

  • MPIX_Comm_ack_failed(MPI_Comm comm, int num_to_ack, int* num_acked) acknowledges that the application intends to ignore the effect of currently known failures on wildcard receive completions and agreement return values (API). See MPIX_Comm_ack_failed.

5.3.3. Supported Systems

There are several MPI communication engines available in Open MPI, notably:

  • PML: ob1, cm, ucx

  • MTL: ofi, portals4, psm2

However, in Open MPI v5.0.5, only ob1 is fully adapted to support fault tolerance. The UCX PML has been successfully tested in some setups, but at this point we cannot confirm that all UCT devices are fully capable to provide the necessary features.

ob1 uses BTL (“Byte Transfer Layer”) components for each supported network. ob1 supports a variety of networks that can be used in combination with each other. Collective operations (blocking and non-blocking) use an optimized implementation on top of ob1.

  • Loopback (send-to-self)

  • TCP

  • UCT (InfiniBand)

  • uGNI (Cray Gemini, Aries)

  • Shared Memory (FT supported with CMA and XPMEM; KNEM is untested)

  • Tuned and non-blocking collective communications

A full list of supported, untested and disabled components is provided later in this document.

5.3.4. ULFM web site

More information (tutorials, examples, build instructions for leading top500 systems) is also available in the Fault Tolerance Research Hub website: https://fault-tolerance.org

5.3.5. Bibliographic References

If you are looking for, or want to cite a general reference for ULFM, please use:

Wesley Bland, Aurelien Bouteiller, Thomas Herault, George Bosilca, Jack J. Dongarra: Post-failure recovery of MPI communication capability: Design and rationale. IJHPCA 27(3): 244-254 (2013).

Available from DOI 10.1177/1094342013488238.

5.3.6. Building ULFM support in Open MPI

In Open MPI v5.0.5, ULFM support is built-in by default — that is, when you build Open MPI, unless you specify --without-ft, ULFM support is automatically available (but is inactive unless enabled at runtime).

Optionally, you can specify --with-ft ulfm to ensure that ULFM support is definitely built.

Note

ULFM Fault Tolerance does not apply to OpenSHMEM. It is recommended that if you are going to use ULFM, you should disable building OpenSHMEM with --disable-oshmem.

5.3.7. Running ULFM Open MPI

5.3.7.1. Building your application

As ULFM is still an extension to the MPI standard, you will need to #include <mpi-ext.h> in C, or use mpi_ext in Fortran to access the supplementary error codes and functions.

Compile your application as usual, using the provided mpicc or mpifort wrappers.

5.3.7.2. Running your application

You can launch your application with fault tolerance by simply using the normal Open MPI mpirun launcher, with the --with-ft ulfm CLI option (or its synonym --with-ft mpi):

shell$ mpirun --with-ft ulfm ...

Important

By default, fault tolerance is not active at run time. It must be enabled via --with-ft ulfm.

5.3.7.3. Running under a batch scheduler

ULFM can operate under a job/batch scheduler, and is tested routinely with ALPS, PBS, and Slurm. One difficulty comes from the fact that many job schedulers handle failures by triggering an immediate “cleanup” of the application as soon as any process fails. In addition, failure detection subsystems integrated into PRTE are not active in direct launch scenarios and may not have a launcher specific alternative. This may cause the application to not detect failures and lock. In order to avoid these problems, it is preferred that you use mpirun within an allocation (e.g., salloc, sbatch, qsub) rather than a direct launch.

  • SLURM is tested and supported with fault tolerance.

    Important

    Use mpirun in an salloc or sbatch allocation. Direct launch with srun is not supported.

  • PBS/Torque is tested and supported with fault tolerance.

    Important

    Use mpirun in a qsub allocation. Direct launch with aprun is not supported.

  • LSF is untested with fault tolerance.

5.3.7.4. Run-time tuning knobs

ULFM comes with a variety of knobs for controlling how it runs. The default parameters are sane and should result in good performance in most cases. You can change the default settings with --mca mpi_ft_foo <value> for Open MPI options, and with --prtemca errmgr_detector_bar <value> for PRTE options.

Important

The main control for enabling/disabling fault tolerance at runtime is the --with-ft ulfm (or its synomym --with-ft mpi) mpirun CLI option. This option sets up multiple subsystems in Open MPI to enable fault tolerance. The options described below are best used to override the default behavior after the --with-ft ulfm opion is used.

5.3.7.4.1. PRTE level options

  • prrte_enable_ft <true|false> (default: false) controls automatic cleanup of apps with failed processes within mpirun. This option is automatically set to true when using --with-ft ulfm.

  • errmgr_detector_priority <int> (default 1005) selects the PRRTE-based failure detector. Only available when prte_enable_recovery is true. You can set this to 0 when using the (experimental) Open MPI detector instead.

  • errmgr_detector_heartbeat_period <float> (default: 5e0) controls the heartbeat period. Recommended value is 1/2 of the timeout.

  • errmgr_detector_heartbeat_timeout <float> (default: 1e1 seconds) heartbeat timeout (i.e. failure detection speed). Recommended value is 2 times the heartbeat period. The default setup is tuned for failure-free performance at the expense of fault detection reactivity. In environments where faults are expected to be common, less conservative values can be used (e.g., 100ms); Values lower than the TCP poll rate (typically 10ms) can cause false positive.

5.3.7.4.2. Open MPI level options

Default values are applied to some Open MPI parameters when using mpirun --with-ft ulfm. These defaults are obtained from the ft-mpi aggregate MCA param file $installdir/share/openmpi/amca-param-sets/ft-mpi. You can tune the runtime behavior of ULFM by either setting or unsetting variables in this file, or by overriding the variable on the command line (e.g., --mca btl ofi,self).

Important

Note that if fault tolerance is disabled at runtime, (that is, when not using --with-ft ulfm), the ft-mpi AMCA param file is not loaded, thus components that are unsafe for fault tolerance will load normally (this may change observed performance when comparing with and without fault tolerance).

  • mpi_ft_enable <true|false> (default: false) permits turning on/off fault tolerance at runtime. This option is automatically set to true from the aggregate MCA param file ft-mpi loaded when using --with-ft ulfm. When false, failure detection is disabled; Interfaces defined by the fault tolerance extensions are substituted with dummy non-fault tolerant implementations (e.g., MPIX_Comm_agree is implemented with MPI_Allreduce); All other controls below become irrelevant.

  • mpi_ft_verbose <int> (default: 0) increases the output of the fault tolerance activities. A value of 1 will report detected failures.

  • mpi_ft_detector <true|false> (default: false), DEPRECATED controls the activation of the Open MPI level failure detector. When this detector is turned off, all failure detection is delegated to PRTE (see above). The Open MPI level fault detector is experimental. There is a tradeoff between failure detection accuracy and performance with this detector. Users that experience accuracy issues may enable a more precise mode. See the tuning knobs below to adjust to taste; The Open MPI failure detector operates on MPI_COMM_WORLD exclusively. Processes connected from MPI_COMM_CONNECT/ACCEPT and MPI_COMM_SPAWN may occasionally not be detected when they fail.

    Caution

    This component is deprecated. Failure detection is now performed at the PRTE level. See the section above on controlling PRTE behavior for information about how to tune the failure detector.

  • mpi_ft_detector_thread <true|false> (default: false) controls the use of a thread to emit and receive failure detector’s heartbeats. Setting this value to “true” will also set MPI_THREAD_MULTIPLE support, which has a noticeable effect on latency (typically 1us increase). You may want to enable this option if you experience false positive processes incorrectly reported as failed with the Open MPI failure detector.

    Important

    This option is only relevant when mpi_ft_detector is true.

  • mpi_ft_detector_period <float> (default: 3e0 seconds) heartbeat period. Recommended value is 1/3 of the timeout. _Values lower than 100us may impart a noticeable effect on latency (typically a 3us increase)._

    Important

    This option is only relevant when mpi_ft_detector is true.

  • mpi_ft_detector_timeout <float> (default: 1e1 seconds) heartbeat timeout (i.e. failure detection speed). Recommended value is 3 times the heartbeat period.

    Important

    This option is only relevant when mpi_ft_detector is true.

5.3.8. Known Limitations in ULFM

  • InfiniBand support is provided through the UCT BTL; fault tolerant operation over the UCX PML is not yet supported for production runs.

  • TOPO, FILE, RMA are not fault tolerant. They are expected to work properly before the occurrence of the first failure.

5.3.9. Modified, Untested and Disabled Components

Frameworks and components are listed below and categorized into one of three classifications:

  1. Modified: This framework/component has been specifically modified such that it will continue to work after a failure.

  2. Untested: This framework/component has not been modified and/or tested with fault tolerance scenarios, and _may_ malfunction after a failure.

  3. Disabled: This framework/component will cause unspecified behavior when fault tolerance is enabled. As a consequence, it will be disabled when the --with-ft ulfm option is used (see above for defails about implicit parameters loaded from the ft-mpi aggregate param file).

Any framework or component not listed below are categorized as Unmodified, meaning that it is unmodified for fault tolerance, but will continue to work correctly after a failure.

  • pml: MPI point-to-point management layer

    • monitoring, v: untested (they have not been modified to handle faults)

    • cm, crcpw, ucx: disabled

  • btl: Point-to-point Byte Transfer Layer

    • ofi, portals4, smcuda, usnic, sm(+knem): untested (they may work properly, please report)

  • mtl: Matching transport layer Used for MPI point-to-point messages on some types of networks

    • All mtl components are disabled

  • coll: MPI collective algorithms

    • cuda, inter, sync, sm: untested (they have not been modified to handle faults, but we expect correct post-fault behavior)

    • hcoll, portals4 disabled (they have not been modified to handle faults, and we expect unspecified post-fault behavior)

  • osc: MPI one-sided communications

    • All osc components are untested (they have not been modified to handle faults, and we expect unspecified post-fault behavior)

  • io: MPI I/O and dependent components

    • fs: File system functions for MPI I/O

    • fbtl: File byte transfer layer: abstraction for individual read/write operations for OMPIO

    • fcoll: Collective read and write operations for MPI I/O

    • sharedfp: Shared file pointer operations for MPI I/O

    • All components in these frameworks are unmodified, untested (we expect clean post-failure abort)

  • vprotocol: Checkpoint/Restart components

    • All vprotocol components are untested

  • threads, wait-sync: Multithreaded wait-synchronization object

    • argotbots, qthreads: disabled (these components have not been modified to handle faults; we expect post-failure deadlock)

5.3.10. Changelog

5.3.10.1. ULFM Integrated in Open MPI

As of v5.0.5, ULFM is now integrated directly in to the community release of Open MPI. The following sections describe previous ULFM standlone releases.

5.3.10.2. ULFM Standalone Release 4.0.2u1

This is a stability and upstream parity upgrade. It is based on the most current Open MPI Release (v4.0.2, October 2019).

  • This release is based on Open MPI release v4.0.2 (ompi #cb5f4e737a).

  • This release is based on ULFM master (ulfm #0e249ca1).

  • New features

    • Support for the UCT BTL enters beta stage.

  • Bugfixes

    • High sensitivity to noise in the failure detector.

    • Deadlocks when revoking while BTL progress threads are updating messages.

    • A case where the failure detector would keep observing a dead process forever.

    • Disable the use of external pmix/libevent by default (the internals are modified to handle error cases).

    • Clean error paths leaving some rdma registration dangling.

    • Do not remove the orte job/proc session dir prematurely upon error.

5.3.10.3. ULFM Standalone Release 4.0.1u1

This is a stability and upstream parity upgrade. It improves stability, performance and is based on the most current Open MPI Release (v4.0.1, May 2019).

  • This release is based on Open MPI release v4.0.1 (ompi #b780667).

  • This release is based on ULFM master (ulfm #cf8dc43f).

  • New features

    • Addition of the MPI_Comm_is_revoked function

    • Renamed ftbasic collective component to ftagree

    • Restored the pcollreq extension

  • Bugfixes

    • Failures of node-local siblings were not always detected

    • Failure propagation and detection was slowed down by trying to notify known dead processes

    • There were deadlocks in multithreaded programs

    • There were issues with PMPI when compiling Fortran Interfaces

    • There were deadlocks on OS-X

5.3.10.4. ULFM Standalone Release 2.1

This release is a bugfix and upstream parity upgrade. It improves stability, performance and is based on the most current Open MPI main (November 2018).

  • ULFM is now based upon Open MPI main branch (#37954b5f).

  • ULFM tuning MCA parameters are exposed by ompi_info.

  • Fortran 90 bindings have been updated

  • Bugfixes:

    • Correct the behavior of process placement during an MPI_COMM_SPAWN when some slots were occcupied by failed processes.

    • MPI_COMM_SPAWN accepts process placement directives in the Info object.

    • Fixed deadlocks in some NBC collective operations.

    • Crashes and deadlocks in MPI_FINALIZE have been resolved.

    • Any-source requests that returned with an error status of MPIX_PROC_FAILED_PENDING can now correctly complete during later MPI_WAIT/TEST.

5.3.10.5. ULFM Standalone Release 2.0

Focus has been toward integration with current Open MPI main (November 2017), performance, and stability.

  • ULFM is now based upon Open MPI main branch (#689f1be9). It will be regularly updated until it will eventually be merged.

  • Fault Tolerance is enabled by default and is controlled with MCA variables.

  • Added support for multithreaded modes (MPI_THREAD_MULTIPLE, etc.)

  • Added support for non-blocking collective operations (NBC).

  • Added support for CMA shared memory transport (Vader).

  • Added support for advanced failure detection at the MPI level. Implements the algorithm described in “Failure detection and propagation in HPC systems,” DOI 10.1109/SC.2016.26.

  • Removed the need for special handling of CID allocation.

  • Non-usable components are automatically removed from the build during configure

  • RMA, FILES, and TOPO components are enabled by default, and usage in a fault tolerant execution warns that they may cause undefined behavior after a failure.

  • Bugfixes:

    • Code cleanup and performance cleanup in non-FT builds; –without-ft at configure time gives an almost stock Open MPI.

    • Code cleanup and performance cleanup in FT builds with FT runtime disabled; –mca ft_enable_mpi false thoroughly disables FT runtime activities.

    • Some error cases would return ERR_PENDING instead of ERR_PROC_FAILED in collective operations.

    • Some test could set ERR_PENDING or ERR_PROC_FAILED instead of ERR_PROC_FAILED_PENDING for ANY_SOURCE receptions.

5.3.10.6. ULFM Standalone Release 1.1

Focus has been toward improving stability, feature coverage for intercomms, and following the updated specification for MPI_ERR_PROC_FAILED_PENDING.

  • Forked from Open MPI 1.5.5 devel branch

  • Addition of the MPI_ERR_PROC_FAILED_PENDING error code, as per newer specification revision. Properly returned from point-to-point, non-blocking ANY_SOURCE operations.

  • Alias MPI_ERR_PROC_FAILED, MPI_ERR_PROC_FAILED_PENDING and MPI_ERR_REVOKED to the corresponding standard blessed -extension- names MPIX_ERR_xxx.

  • Support for Intercommunicators:

    • Support for the blocking version of the agreement, MPI_COMM_AGREE on Intercommunicators.

    • MPI_COMM_REVOKE tested on intercommunicators.

  • Disabled completely (.ompi_ignore) many untested components.

  • Changed the default ORTE failure notification propagation aggregation delay from 1s to 25ms.

  • Added an Open MPI internal failure propagator; failure propagation between SM domains is now immediate.

  • Bugfixes:

    • SendRecv would not always report MPI_ERR_PROC_FAILED correctly.

    • SendRecv could incorrectly update the status with errors pertaining to the Send portion of the Sendrecv.

    • Revoked send operations are now always completed or remote cancelled and may not deadlock anymore.

    • Cancelled send operations to a dead peer will not trigger an assert when the BTL reports that same failure.

    • Repeat calls to operations returning MPI_ERR_PROC_FAILED will eventually return MPI_ERR_REVOKED when another process revokes the communicator.

5.3.10.7. ULFM Standalone Release 1.0

Focus has been toward improving performance, both before and after the occurrence of failures. The list of new features includes:

  • Support for the non-blocking version of the agreement, MPI_COMM_IAGREE.

  • Compliance with the latest ULFM specification draft. In particular, the MPI_COMM_(I)AGREE semantic has changed.

  • New algorithm to perform agreements, with a truly logarithmic complexity in number of ranks, which translates into huge performance boosts in MPI_COMM_(I)AGREE and MPI_COMM_SHRINK.

  • New algorithm to perform communicator revocation. MPI_COMM_REVOKE performs a reliable broadcast with a fixed maximum output degree, which scales logarithmically with the number of ranks.

  • Improved support for our traditional network layer:

    • TCP: fully tested

    • SM: fully tested (with the exception of XPMEM, which remains unsupported)

  • Added support for High Performance networks

    • Open IB: reasonably tested

    • uGNI: reasonably tested

  • The tuned collective module is now enabled by default (reasonably tested), expect a huge performance boost compared to the former basic default setting

    • Back-ported PBS/ALPS fixes from Open MPI

    • Back-ported OpenIB bug/performance fixes from Open MPI

    • Improve Context ID allocation algorithm to reduce overheads of Shrink

    • Miscellaneous bug fixes

5.3.10.8. Binary Compatibility

ULFM Open MPI is binary compatible with any version of Open MPI compatible with the underlying Open MPI main branch or release (see the binary compatibility and version number section in the upstream Open MPI README). That is, applications compiled with a compatible Open MPI can run with the ULFM Open MPI mpirun and MPI libraries. Conversely, as long as the application does not employ one of the MPIX functions, which are exclusively defined in ULFM Open MPI, an application compiled with ULFM Open MPI can be launched with a compatible Open MPI mpirun and run with the non-fault tolerant MPI library.

5.3.11. Contacting the Authors

Found a bug? Got a question? Want to make a suggestion? Want to contribute to ULFM Open MPI? Working on a cool use-case? Please let us know!

The best way to report bugs, send comments, or ask questions is to sign up on the user’s mailing list: ulfm+subscribe@googlegroups.com

Because of spam, only subscribers are allowed to post to these lists (ensure that you subscribe with and post from exactly the same e-mail address — joe@example.com is considered different than joe@mycomputer.example.com!). Visit these pages to subscribe to the lists: https://groups.google.com/forum/#!forum/ulfm

When submitting questions and problems, be sure to include as much extra information as possible. See the Getting help section for more details.

Thanks for your time.