<div dir="auto">Got it. <div dir="auto">Thank you, Giuseppe and Ken.</div><div dir="auto"><br></div><div dir="auto">Regards,</div><div dir="auto">Xiaopeng</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Thu, Jun 6, 2019, 2:56 PM Congiu, Giuseppe <<a href="mailto:gcongiu@anl.gov">gcongiu@anl.gov</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Actually, the fix uses a combination of hostname and random number, which seed is a time stamp. <br>
I don’t remember why exactly we didn’t go for the hostname only but I suspect it is because this might not be<br>
unique. Adding the random number with a timestamp seed should be robust enough against collisions.<br>
<br>
Giuseppe<br>
<br>
> On Jun 6, 2019, at 8:25 AM, Raffenetti, Kenneth J. via discuss <<a href="mailto:discuss@mpich.org" target="_blank" rel="noreferrer">discuss@mpich.org</a>> wrote:<br>
> <br>
> On 6/6/19 12:52 AM, Xiaopeng Duan wrote:<br>
>> Thank you, Ken.<br>
>> <br>
>> We were having another problem with 3.3, and will try it once we fixed <br>
>> our issue.<br>
>> <br>
>> Just my couriosity, why a random number was chosen for the fix instead <br>
>> of hostname or address? Looks to me the random number still has some <br>
>> possibility to repeat (although very rare), but hostnames and addresses <br>
>> should be unique in a system.<br>
> <br>
> I had the same thought when looking back at this patch. Maybe Giuseppe <br>
> can share why that was added. I'm fairly sure it can be safely removed.<br>
> <br>
> Ken<br>
> <br>
>> <br>
>> Regards,<br>
>> Xiaopeng<br>
>> <br>
>> On Wed, Jun 5, 2019, 8:29 AM Raffenetti, Kenneth J. <br>
>> <<a href="mailto:raffenet@mcs.anl.gov" target="_blank" rel="noreferrer">raffenet@mcs.anl.gov</a> <mailto:<a href="mailto:raffenet@mcs.anl.gov" target="_blank" rel="noreferrer">raffenet@mcs.anl.gov</a>>> wrote:<br>
>> <br>
>> We added a similar fix in <a href="https://github.com/pmodels/mpich/pull/2788" rel="noreferrer noreferrer" target="_blank">https://github.com/pmodels/mpich/pull/2788</a>.<br>
>> This was included in the MPICH 3.3 release.<br>
>> <br>
>> Ken<br>
>> <br>
>> On 6/4/19 11:27 PM, Xiaopeng Duan via discuss wrote:<br>
>>> Hi, MPICH experts,<br>
>>> <br>
>>> We are working on a dynamic master-worker flow using<br>
>>> mpi_comm_connect/mpi_com_accept. In some cases when the total<br>
>> number of<br>
>>> worker process is large, they may get the same kvs_name and<br>
>> confuse the<br>
>>> internal group identifiers. This was traced to the naming<br>
>> convention in<br>
>>> HYD_pmcd_pmi_allocate_kvs() that considers only process id, while<br>
>> two<br>
>>> processes on different machines may have the same pid. I tried to<br>
>> add<br>
>>> host name (from unistd.h>gethostname) to the name, i.e.<br>
>>> 'kvs_HOSTNAME_PID_pgid', then everything is working fine in our<br>
>> testing.<br>
>>> <br>
>>> So I'm wondering if this change is safe (we may need it for our<br>
>> release)<br>
>>> and if it would go into the official MPICH release some time.<br>
>>> <br>
>>> Thank you very much.<br>
>>> <br>
>>> Xiaopeng<br>
>>> <br>
>>> _______________________________________________<br>
>>> discuss mailing list <a href="mailto:discuss@mpich.org" target="_blank" rel="noreferrer">discuss@mpich.org</a> <mailto:<a href="mailto:discuss@mpich.org" target="_blank" rel="noreferrer">discuss@mpich.org</a>><br>
>>> To manage subscription options or unsubscribe:<br>
>>> <a href="https://lists.mpich.org/mailman/listinfo/discuss" rel="noreferrer noreferrer" target="_blank">https://lists.mpich.org/mailman/listinfo/discuss</a><br>
>>> <br>
>> <br>
> _______________________________________________<br>
> discuss mailing list <a href="mailto:discuss@mpich.org" target="_blank" rel="noreferrer">discuss@mpich.org</a><br>
> To manage subscription options or unsubscribe:<br>
> <a href="https://lists.mpich.org/mailman/listinfo/discuss" rel="noreferrer noreferrer" target="_blank">https://lists.mpich.org/mailman/listinfo/discuss</a><br>
<br>
</blockquote></div>