Dear all,<br>I would like to know if there is some trick to use parallel MPI/IO in HDF5 file on Lustre?<br>I have been trying for while to fix the issue that happens in Lustre.<br><br>Basically, I have 8 process writing small datas in the same dataset in an attempt.
Each attempt write differentdatathan before, and when an attempt comes close to 2500, the I/O parallel process broken. There is too a anotherserial process writing otherdatain the same file<br><br><br><div>Ifound that is acommonLustre issue which<span style="font-family:sans-serif;font-size:13px;line-height:19px;background-color:rgb(255,255,255)">do not support locking! So I tried this parameters above, but my application</span><span style="background-color:rgb(255,255,255);font-family:sans-serif;font-size:13px;line-height:19px">still</span><span style="background-color:rgb(255,255,255);font-family:sans-serif;font-size:13px;line-height:19px"></span><span style="background-color:rgb(255,255,255);font-family:sans-serif;font-size:13px;line-height:19px">broken.. Also,</span>I read in somewhere to set the "stripesize" to count 8 (number of OST) for parallel IO, but stilldoesn'twoks.</div>
<div><pre style="font-family:monospace,Courier;padding:1em;border:1px dashed rgb(47,111,171);background-color:rgb(249,249,249);line-height:1.1em;font-size:13px"> MPI_Info_create(&info);
/* Disables ROMIO's data-sieving */
MPI_Info_set(info, "romio_ds_read", "disable");
MPI_Info_set(info, "romio_ds_write", "disable");
/* Enable ROMIO's collective buffering */
MPI_Info_set(info, "romio_cb_read", "enable");
<span style="white-space:normal">MPI_Info_set(info, "romio_cb_write", "enable");</span>
</pre><span style="background-color:rgb(249,249,249);font-family:monospace,Courier;font-size:13px;line-height:1.1em"></span>
<a href="https://wickie.hlrs.de/platforms/index.php/MPI-IO">https://wickie.hlrs.de/platforms/index.php/MPI-IO</a><div><span style="background-color:rgb(249,249,249)"><br></span></div><div>Iranmy application on PVFS, and works fine. I know that unlike the Lustre, PVFS use non-alligned striping data. Could be it the reason that works in PVFS?<br>
<br>Regards<br clear="all"><div><br></div>-- <br>Tiago Steinmetz Soares<div>MSc Student of Computer Science - UFSC</div><br>
</div></div>