Error writing ndb.QP file during evGWo

[Bruno]

Hello!

I'm having problems running the evGWo calculation. The calculation runs fine until it ends the G3Wo and then it gives me the following error at the end of the log files l- :

<02h-22m> P1: [07.05] QP properties and I/O
P1: [ERROR] STOP signal received while in[07.05] QP properties and I/O
P1: [ERROR] Writing File ./gw_final//ndb.QP; Variable NOT DEFINED; Permission denied

Before that everything was being written fine. In anycase a tried to submit this calculation again and I got the exact same error also at the end of G3Wo. Is it something related to the yambo regarding maybe the file size? Or is it possible that it is something cluster related? I've never received this permission denied error before in yambo calculations, but this is my first evGWo one, so I suppose it is related to it somehow.

Regards,

[claudio]

Dear Bruno

I will check the code to see if there is a problem.
In the while I advice you to make a script and use the procedure described here

https://www.attaccalite.com/self-consis ... ambo-code/

for the self-consistent GW on G or on G and W.
The advantage of using an external script is that you can perform a restart, not implemented in the fortran loop of yambo.

best
Claudio

[Bruno]

Hello!

Thank you very much for the suggestion Claudio, I'm doing that right now.

Regards,

[Bruno]

Hello!

Just to make sure of one thing, during the actual evGW0 cycle the code does not recompute the EXS term right? Just the G0W0, G1W0... is that right?

Doing it by the mentioned manual way is it possible to prevent it to recompute this term? Because it takes quite a while. There is any file that I should be copying to another folder?

Regards,

[Daniele Varsano]

Dear Bruno,
you can copy the ndb.HF* database in the SAVE directory, this will avoid to recompute the HF term.

Best,
Daniele

[Bruno]

Dear Daniele,

Thank you very much!

Regards,

[Christian Koenig]

Hi All,

Not sure if this issue is already solved, but I get a very similar issue.

I compiled version 5.0.0 on two different machines. On the first one,
as far as I can tell, the code seems to run without problems for G0W0.
Unfortunately, I don't have access to this machine in the moment
and I can't verify if the calculation finished.
On the second machine, compiling works just as for the older versions.
However, p2y will crash with

***************************** p2y output **************************
<---> DBs path set to : .
<---> detected QE data format : qexsd
<---> == PWscf v.6.x generated data (QEXSD fmt) ==
<---> Header/K-points/Energies... done
<03s> Cell data...
... the usual stuff here ...
<05s> == DB1 (Gvecs and more) ...
[ERROR] Writing File ./SAVE//ns.db1; Variable NOT DEFINED; Permission denied
***************************** p2y output **************************

The same happens in the next step for "yambo" (without options)
if I generate the database with a previous version.

Note that there are two spaces between "Variable" and "NOT DEFINED",
so I assume that some variable's name is an empty string.
Could that be a netcdf problem?

Any help would be appreciated.


Best,

Christian

[andrea.ferretti]

Dear Christian,

thanks for reporting... this is quite puzzling. This could be surely related to some compilation issue somehow related to or affecting netcdf.
Alternatively, a variable name getting empty ?

In order to help it would be useful if you could send the config.log file (where one should also be able to understand a bit about the underlying machine)

take care
Andrea

[Christian Koenig]

Dear Andrea,

Thanks for the response. I figured out that the problem is indeed caused by netcdf and I can now run the code on both machines.

When compiling yambo version 5.0.0, netcdf v4 is used by default. My old binaries were linked to v3.
I also observed that during "make all" a warning appeared that a shared library (of netcdf) was linked against a static library.
Not sure if this was specific to me or if this is a problem in the Makefile if those libraries are not pre-installed.
Thus, either v3 and v4 of netcdf were not compatible here - or the linker problem caused the crash at runtime.

The solution for me was to use the option "--enable-netcdf-v3" with configure.
Hopefully this helps if someone encounters the same problem again.


Best regards

Christian

[Davide Sangalli]

Just a comment.
NETCDF v4 library is fully capable to read files created with NETCDF v3.
The other way around does not work.

You can compile yambo with v3, but yambo_rt will not work.
Thus it is suggested to use v4.

Best,
D.