try to recover from an OOM during citation index reading and fail-over

to second solr core in case of unrecoverable OOM.

source/net/yacy/kelondro/blob/HeapReader.java

@@ -539,8 +539,14 @@ public class HeapReader {
             try {
                 blob = new byte[len];
             } catch (final OutOfMemoryError e) {
+                // try once again after GC
+                MemoryControl.gc(1000, "HeapReader.get()/blob");
+                try {
+                    blob = new byte[len];
+                } catch (final OutOfMemoryError ee) {
                     throw new SpaceExceededException(len, "HeapReader.get()/blob");
                 }
+            }
             this.file.readFully(blob, 0, blob.length);
 
             return blob;

source/net/yacy/search/index/Segment.java

@@ -360,6 +360,7 @@ public class Segment {
             this.externalHosts = new RowHandleSet(6, Base64Order.enhancedCoder, 0);
             this.internalIDs = new RowHandleSet(12, Base64Order.enhancedCoder, 0);
             this.externalIDs = new RowHandleSet(12, Base64Order.enhancedCoder, 0);
+            try {
                 if (connectedCitation()) {
                     // read the references from the citation index
                     ReferenceContainer<CitationReference> references;
@@ -379,6 +380,10 @@ public class Segment {
                         }
                     }
                 }
+            } catch (SpaceExceededException e) {
+                // the Citation Index got too large, we ignore the problem and hope that a second solr index is attached which will take over now
+                if (Segment.this.fulltext.writeToWebgraph()) internalIDs.clear();
+            }
             if ((internalIDs.size() == 0 || !connectedCitation()) && Segment.this.fulltext.writeToWebgraph()) {
                 // reqd the references from the webgraph
                 SolrConnector webgraph = Segment.this.fulltext.getWebgraphConnector();