Your request generated a server error, HTTP code 500

[pdi-uk]

Time Out Error from HAPI FHIR Server (Request timed out after 60494ms) causes the script to terminate.

Is it possible to increase the number of trials of sending requests to the FHIR server?
Is it possible to decrease the size of the requests?

[palmjulia]

I think @NandhiniS08 pushed an image of the docker container with reduced request sizes to https://hub.docker.com/r/nandhinis08/projectathon6-miracum1 because there was a similar problem in Leipzig (see MII slack), but we don't know yet if this helped..

Is there a chance you can try this image? If you want to build yourself maybe @NandhiniS08 can push the corresponding R code to a seperate branch for you to check out.

[pdi-uk]

Yes. I would prefer to run the R code directly.

[palmjulia]

Okay if you are comfortable with running R directly then maybe you'll be quicker applying the change yourself. In Line 352 youd have to change nchar_for_ids <- 1800 - (nchar(conf$serverbase)+nchar_loincs) to nchar_for_ids <- 900 - (nchar(conf$serverbase)+nchar_loincs).
This will reduce the number of Patients per request for which the LOINC coded Observations are downloaded (because the total request length is reduced and thus the Patients will be distributed over a higher number of requests, with fewer Patients per request). In Leipzig this change seemed to fix the timeout problem.

[mematt]

Thank you @palmjulia and @NandhiniS08 for the solutions.
We have just received the same HTTP 500 error from Wuerzburg.
It seems to come up quite frequently, thus I will add the quick fix to the readme right away.

[palmjulia]

Yes I hope it is actually the same problem in all these places. In Jena (Smile Server) the original script ran flawlessly but the Smile is probably a bit stronger then the Hapi.

[pdi-uk]

I remember trying the same fix with the last version.

[palmjulia]

Probably similar, yes, because back then the loinc codes were not taken into account in the url length if I remember correctly. But now they already are and the computational load is still to high, or so it seems. So further (quite drastic) reduction might help...

[pdi-uk]

I decreased to
nchar_for_ids <- 500 - (nchar(conf$serverbase)+nchar_loincs)
and still get the error.

[NandhiniS08]

Decreasing to 500 will not help because of the fact that the length may go to a negative number
nchar_loincs = 616. Can you see if 900 works for you?

[pdi-uk]

900 did not work either

[NandhiniS08]

Oh Thats not good! May I ask if this error comes while execution of the observation request?

[pdi-uk]

No, while executing the Condition request:

Starting download of ALL! bundles of resource type Condition from FHIR base URL http://as13-diz:18989/fhir.

This may take a while...
Fehler in check_response(response = response, log_errors = log_errors) :
Your request generated a server error, HTTP code 500. For more information see the generated error file.
Ruft auf: lapply -> FUN -> fhir_search -> get_bundle -> check_response
Ausführung angehalten

Are the ressources downloaded using a POST or GET request?
For the POLAR scripts there where similar problems using POST requests for the condition ressources, but not for encounter.

[pdi-uk]

HAPI Version 5.3.0

[pdi-uk]

In Postman the last next-link was resolving normally.

[pdi-uk]

I tried to change
max_bundles = 100 -> max_bundles = 40
in line 131
now the script runs successfully.