Fix basic auth polling to skip retries when cachedSerializedGroupMappingMap returns 404 (#9354)

This commit is contained in:
Jonathan Wei 2020-02-12 16:52:03 -08:00 committed by GitHub
parent 31528bcdaf
commit 48a0681f7e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 10 additions and 0 deletions

View File

@ -51,6 +51,7 @@ import org.apache.druid.security.basic.authorization.entity.UserAndRoleMap;
import org.apache.druid.server.security.Authorizer;
import org.apache.druid.server.security.AuthorizerMapper;
import org.jboss.netty.handler.codec.http.HttpMethod;
import org.jboss.netty.handler.codec.http.HttpResponseStatus;
import org.joda.time.Duration;
import javax.annotation.Nullable;
@ -418,6 +419,15 @@ public class CoordinatorPollingBasicAuthorizerCacheManager implements BasicAutho
req,
new BytesFullResponseHandler()
);
// cachedSerializedGroupMappingMap is a new endpoint introduced in Druid 0.17.0. For backwards compatibility, if we
// get a 404 from the coordinator we stop retrying. This can happen during a rolling upgrade when a process
// running 0.17.0+ tries to access this endpoint on an older coordinator.
if (responseHolder.getStatus().equals(HttpResponseStatus.NOT_FOUND)) {
LOG.warn("cachedSerializedGroupMappingMap is not available from the coordinator, skipping fetch of group mappings for now.");
return null;
}
byte[] groupRoleMapBytes = responseHolder.getContent();
GroupMappingAndRoleMap groupMappingAndRoleMap = objectMapper.readValue(