Updates to the canonical validation resource cache in VersionSpecificWorkerContextWrapper
#6506
Conversation
… that canonical CodeSystem resources are also cached. Use the terminology troubleshooting logs throughout the validation module.
|
Formatting check succeeded! |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #6506 +/- ##
============================================
- Coverage 83.54% 83.46% -0.09%
- Complexity 27432 27882 +450
============================================
Files 1707 1752 +45
Lines 106185 108083 +1898
Branches 13397 13563 +166
============================================
+ Hits 88710 90208 +1498
- Misses 11750 12045 +295
- Partials 5725 5830 +105 ☔ View full report in Codecov by Sentry. |
|
FYI @codeforgreen I believe this PR should be redundant once this merges (it also fixes the same issue by just never expiring StructureDefinions): #6508 |
| would be returned which suggest the profile was not applied properly or used. | ||
| That happens because the validation cache entry expired within the $validate call | ||
| and the FHIR Core Validation library expects the same object instance for the same profile to be returned. | ||
| This has been fixed by increasing the cache expiry to 10min." |
There was a problem hiding this comment.
question: Does this mean we are still vulnerable to rare failures if a cache entry expires in the middle of processing?
| ourLog.warn( | ||
| String.format("Multiple results found for URL '%s', only the first will be considered.", url)); | ||
| if (results != null && !results.isEmpty()) { | ||
| if (results.size() > 1 && ourLog.isWarnEnabled()) { |
There was a problem hiding this comment.
trivial: the ourLog.isWarnEnabled() check seems redundant since the call to warn uses a format string.
|
|
||
| import java.io.IOException; | ||
|
|
||
| public class DefaultProfileValidationSupportNpmStrategy extends NpmPackageValidationSupport { | ||
| private static final Logger ourLog = LoggerFactory.getLogger(DefaultProfileValidationSupportNpmStrategy.class); | ||
| private static final Logger ourLog = Logs.getTerminologyTroubleshootingLog(); |
There was a problem hiding this comment.
Nice to see a new troubleshooting log.
…ors-profile-cache-expiry
Fixes #6505
Please note that the fix will be backported to version 7.4 as well.
What was done:
CacheTimeouts).VersionSpecificWorkerContextWrappersuch that all methods that retrieve validation resources go through the cache using same method.CodeSystemresources in particular was not using this cache and it now is using it.IValidationSupportto useLogs.getTerminologyTroubleshootingLog()VersionSpecificWorkerContextWrapperrelated to the cache and also testing the methods to fetch validation resources at different levels (unit and integration).While working on this issue, some opportunity for optimization was observed which will be tackled in a different issue. Seems that the package installer does not generate snapshots properly. Some snapshots are not saved to the database and since the internal cache does not store generated snapshots when it encounters such cases, that results in some validate calls to be slow because they generate the missing requires snapshots every time the cache entry for the profile expires.