<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<style>body{font-family:Helvetica,Arial;font-size:12px}</style>
</head>
<body style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;">
<div id="bloop_customfont" style="font-family:Helvetica,Arial;font-size:12px; color: rgba(0,0,0,1.0); margin: 0px; line-height: auto;">
The notes from the meeting are now posted:</div>
<div id="bloop_customfont" style="font-family:Helvetica,Arial;font-size:12px; color: rgba(0,0,0,1.0); margin: 0px; line-height: auto;">
<br>
</div>
<div class="bloop_container">
<div class="bloop_frame"><a href="https://github.com/mpiwg-ft/ft-issues/wiki/2017-10-18">https://github.com/mpiwg-ft/ft-issues/wiki/2017-10-18</a>
</div>
</div>
<br>
<div id="bloop_sign_1508435923234936832" class="bloop_sign">
<div>Thanks,</div>
<div>Wesley</div>
<div><br>
</div>
<div>===</div>
<div><br>
</div>
<div>
<div>### Attendees</div>
<div><br>
</div>
<div>* Intel - Wesley, Jim</div>
<div>* UTK - Aurelien</div>
<div>* ORNL - Geoffroy</div>
<div>* Argonne - Yanfei, Ken</div>
<div>* Sandia - Keita</div>
<div>* LLNL - Murali, Ignacio</div>
<div><br>
</div>
<div>## RMA Fault Tolerance (Data Resilience)</div>
<div><br>
</div>
<div>[Link to pull request](https://github.com/mpiwg-ft/mpi-standard/pull/4)</div>
<div><br>
</div>
<div>Continued discussion of whether this work is useful.</div>
<div><br>
</div>
<div>* Aurelien - The description of the failure model is unclear. We need to better differentiate between `MPI_ERR_PROC_FAILED` and `MPI_ERR_DATA_UNAVAILABLE`.</div>
<div>* Jim - Should `MPI_ERR_DATA_UNAVAILABLE` be usable outside of RMA? Does it apply to point-to-point or collectives?</div>
<div>* Jim/Aurelien - Is the justification for this work that flush doesn't allow detection of process failure? They're still not convinced that this is true.</div>
<div> * As long as flush can complete successfully, do we really need to tell you if a process failed on the other end?</div>
<div>* Jim - On the other hand, it might be true that we can't guarantee any process failure detection in _any_ RMA operation. Maybe we should just not allow process failure errors (as opposed to "upgrading" other types of errors to process failure).</div>
<div>* Jim - One place this still makes sense as is is having a process with data corrupted because another process failed during a put. If a third process is reading the bad memory, it _could_ get `MPI_ERR_DATA_UNAVAILABLE` instead of `MPI_ERR_PROC_FAILED`.</div>
<div><br>
</div>
<div>### Bottom Line</div>
<div><br>
</div>
<div>* We're still unclear on the failure model expected here. We probably need to get more feedback from Jeff. </div>
<div>* We also aren't convinced that process failure semantics aren't sufficient to tell the user all of the actionable information that they need.</div>
<div><br>
</div>
<div>## For next week:</div>
<div>* Get feedback from Jeff when he comes back from paternity leave.</div>
<div><br>
</div>
<div>## In the future:</div>
<div>* Start discussing text for FA-MPI and Reinit.</div>
</div>
</div>
</body>
</html>