<meta http-equiv="Content-Type" content="text/html; charset=utf-8"><div dir="ltr">Dear <span class="" id=":70f.1" tabindex="-1">MPICH</span>.<div>I'm using <span class="" id=":70f.2" tabindex="-1">mpich</span> 3.1 (hydra+<span class="" id=":70f.3" tabindex="-1">MPI</span>).</div><div>I execute main application (Main) which calls <span class="" id=":70f.4" tabindex="-1">mpiexec</span>.hydra in following way:</div><div><br></div><div><span class="" id=":70f.5" tabindex="-1">mpiexec</span>.hydra -<span class="" id=":70f.6" tabindex="-1">genvall</span>  -disable-auto-cleanup  -f <span class="" id=":70f.7" tabindex="-1">MpiConfigMachines</span>.<span class="" id=":70f.8" tabindex="-1">txt</span> -launcher=ssh -n 3 <span class="" id=":70f.9" tabindex="-1">MPI</span>_<span class="" id=":70f.10" tabindex="-1">Prog</span> <br></div><div><br></div><div><span class="" id=":70f.11" tabindex="-1">MpiConfigMachines</span>.<span class="" id=":70f.12" tabindex="-1">txt</span> content:<br></div><div><div><a href="http://10.3.2.100:1">10.3.2.100:1</a></div><div><a href="http://10.3.2.101:2">10.3.2.101:2</a></div></div><div><br></div><div>Where 10.3.2.100 is a local host.</div><div>As result I get</div><div><ul><li>Main + single <span class="" id=":70f.13" tabindex="-1">MPI</span>_<span class="" id=":70f.14" tabindex="-1">Prog</span> processes on local computer<br></li><li>2 <span class="" id=":70f.15" tabindex="-1">MPI</span>_<span class="" id=":70f.16" tabindex="-1">Prog</span> processes on remote one.</li></ul><div>Main application establish <span class="" id=":70f.17" tabindex="-1">TCP</span> socket with local <span class="" id=":70f.18" tabindex="-1">MPI</span>_<span class="" id=":70f.19" tabindex="-1">Prog</span>.</div></div><div>Main application establish <span class="" id=":70f.20" tabindex="-1">TCP</span> socket with controller on other computer 10.3.2.170, which is not included in <span class="" id=":70f.21" tabindex="-1">MpiConfigMachines</span>.<span class="" id=":70f.22" tabindex="-1">txt</span> file.</div><div><br></div><div>After executing some time (hours, sometimes days) I see via <span class="" id=":70f.23" tabindex="-1">netstat</span> that created new connection from <span class="" id=":70f.24" tabindex="-1">mpiexec</span>.hydra and controller. </div><div><br></div><div>Before executing <span class="" id=":70f.25" tabindex="-1">mpiexec</span>.hydra I set environment variable</div><div><p class="MsoNormal"><span class="" id=":70f.26" tabindex="-1">setenv</span> <span class="" id=":70f.27" tabindex="-1">MPIEXEC</span>_PORT_RANGE 50010:65535</p><p class="MsoNormal">According to manual this variable limits hydra destination ports to [50010:65535].</p><p class="MsoNormal"><br></p><p class="MsoNormal">I see that hydra uses these ports with <span class="" id=":70f.28" tabindex="-1">MPI</span>_<span class="" id=":70f.29" tabindex="-1">Prog</span>, but connection with controller done on port 701 (controller computer).</p><p class="MsoNormal"><br></p><p class="MsoNormal">Controller program is a server. It can accept connections only.<br></p><p class="MsoNormal"><br></p><p class="MsoNormal">Can you please advice how to stand with this problem?</p><p class="MsoNormal">How hydra recognizes controller <span class="" id=":70f.30" tabindex="-1">IP</span> and establish connection with it?</p><p class="MsoNormal"><br></p><p class="MsoNormal">Sincerely,</p><p class="MsoNormal"><span class="" id=":70f.31" tabindex="-1">Anatoly</span>.</p></div><div><br></div></div>