Strange issue while parsing with aiida-nwchem module

[Nathan Keilbart]

Hello,

I have a personalized fork of aiida-nwchem to make workflows. I am currently running aiida v2.2.2. The issue I am seeing is that after including a new functionality to my workflow, a frequency calculation, it sometimes will go all the way through the workflow giving a finished state and other times will crash at the newly included functionality saying:

05/05/2023 11:51:27 AM <3201582> aiida.orm.nodes.process.workflow.workchain.WorkChainNode: [REPORT] [976976|NwchemBaseWorkChain|results]: required output 'output_parameters' was not an output of NwchemCalculation<976977> (or an incorrect class/output is being exposed).

Like I said, it works sometimes and I can literally rerun the exact same script back to back and get different results. I currently have caching enabled for nwchem calculations as well to help accelerate the testing.

Some things I have attempted to do to remedy the situation is making sure I don't have any rogue daemons running else where. I haven't been able to detect them at this point. I do a full restart of the daemon doing reentry scan; verdi daemon restart --reset. Additionally, I asked to have the RabbitMQ service restarted as initially I was running things on a server that had been upgraded from TOSS3 to TOSS4 and wasn't sure if this caused issues. I have installed in a new python environment as well.

If there are any thoughts of what might be causing this or if I can provide more information please let me know. Thanks.

Nathan

[Sebastiaan Huber]

Hi Nathan,

Sorry for the late reply, but it is indeed a weird situation so required some thought.
The error the `NwchemBaseWorkChain` comes from the fact that the final calculation (i.e. NwchemCalculation<976977>) didn't have the required output node `output_parameters`.
This could be due to caching, although I don't see how that would explain the fact that it is transient.

To check whether it could be related to caching, please check whether calc 976977 was taken from cache, i.e., check whether `load_node(976977).base.caching.get_cache_source()` returns an identifier.
If so, it was cached from that node and you should check if that calculation is actually "complete", i.e., does it actually have the `output_parameters` output.
If not, then this is incorrectly considered to be a "valid" cache source.
To stop it from being considered a cache source, you can either delete the node entirely, or if you need to keep it, you can run `load_node().is_valid_cache = False`.

Just as another note, as of v2.0 the command `reentry scan` is obsolete, and you never need to call it anymore, not even when you install new or change plugins.
You do still have to `verdi daemon restart --reset` when changing plugins though.

Hope that helps,

Sebastiaan

--
AiiDA is supported by the NCCR MARVEL (http://nccr-marvel.ch/), funded by the Swiss National Science Foundation, and by the European H2020 MaX Centre of Excellence (http://www.max-centre.eu/).
 
Before posting your first question, please see the posting guidelines at http://www.aiida.net/?page_id=356 .
---
You received this message because you are subscribed to the Google Groups "aiidausers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to aiidausers+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/aiidausers/e888ce26-a385-41b7-af4c-44bdad2c533cn%40googlegroups.com.

[Nathan Keilbart]

Hi Sebastiaan,

I really appreciate the help on this. It does seem that it was finding multiple entries that were listed as possible sources and one of them that it primarily was reading from didn't finish correctly. After changing the .is_valid_cache option to False it found another one and was finishing a couple of times. One thing I saw though is that it wasn't always finding the other source and was attempting to recompute the value again but since it's dependent on another calculation that the values had been deleted from the server it wasn't able to finish.

I'm not sure how it's searching for a valid caching source but what would you recommend is the best way to start from a clean slate? I disabled caching and am running another set of simulations and hoping that will fix the issue but I'm not sure if it looks for the latest set of calculations that match or something else.

Thanks for updating me on reentry scan. I just had an alias setup and had been using that.

Nathan