Postgres error when saving Observation
# supportb
brendankeeler
01/10/2024, 7:57 PMNot sure if adding here or via Github issues is more appropriate but we are seeing some errors logged due to failures when saving Observations.
Looking at the error:
index row size 4200 exceeds btree version 4 maximum 2704 for index "Observation_valueString_idx"
This means the maximum size of a single index entry was exceeded, so it seems we pulled an Observation with a really big string in the Observation.value and the Medplum database couldn't index it.
Wanted to flag this as it could happen more broadly.
r
rahul1
01/11/2024, 2:32 AM@brendankeeler thanks for the heads up. Could you share an example of the Observation in question? At first glance, It seems like having such a large Observation might be a mistake, bit not sure about the use case
rahul1
01/11/2024, 2:32 AMFiling a GH issue with an example data would be good too - the eng team could take a look to see if there's any workarounds.
rahul1
01/11/2024, 2:35 AMThis is happening because we index all search parameters. Truncation might make sense here, as I can't imagine you'd want to search on the full text string?
b
brendankeeler
01/11/2024, 4:49 AMI deidentified it - it's a very long string (which that field is allowed to be).
brendankeeler
01/11/2024, 4:50 AMYeah, I'm not sure - might be a good question for the FHIR community on best behavior there
brendankeeler
01/11/2024, 4:51 AMI'm not particularly interested in searching using code-value-string, just trying to ensure we don't drop a bunch of lab results. Truncation seems like a good simple fix
brendankeeler
01/11/2024, 4:52 AMI can make a Github issue tomorrow
r
rahul1
01/12/2024, 5:00 AMThank you very much. Out of curiosity, do you know code
PXNb
brendankeeler
01/23/2024, 8:51 PMI don't - the OID suggests it's an Epic specific concept, since all Epic OIDs start with prefix 1.2.840.114350
164 Views