Under low memory conditions, we're catching the OutOfMemory exception in the AbstractReloadingMetadataResolver class to log it, but it's only logged on DEBUG.
We catch the exception it rethrows in the refresh thread, but that doesn't get logged, so all the log shows on INFO is a stream of Next refresh time is... messages that make it appear to be checking for changes but not finding any.
We should also review the service reload code for any similar issue.
To work around this issue prior to the release of V3.2, add the following logger to /opt/shibboleth-idp/conf/logback.xml:
<!-- the following logger works around issue https://issues.shibboleth.net/jira/browse/OSJ-125 -->
<logger name="org.opensaml.saml.metadata.resolver.impl.AbstractReloadingMetadataResolver" level="DEBUG"/>
Tom Zeller
September 15, 2015 at 2:29 PM
Agree, just wanted to make the suggestion.
Scott Cantor
September 15, 2015 at 2:27 PM
I don't think "Out of memory" will be confusing to people (well, not the people who can be helped at all).
Tom Zeller
September 15, 2015 at 2:25 PM
Case closed, just wondering if we should try and catch the OutOfMemoryError to log something "more meaningful", not sure exactly what that would be.
Under low memory conditions, we're catching the OutOfMemory exception in the AbstractReloadingMetadataResolver class to log it, but it's only logged on DEBUG.
We catch the exception it rethrows in the refresh thread, but that doesn't get logged, so all the log shows on INFO is a stream of Next refresh time is... messages that make it appear to be checking for changes but not finding any.
We should also review the service reload code for any similar issue.