Hi, after reading some Solr source code, I might have found the cause:
There was indeed a change in Solr 8.6 that leads to the NullPointerException for the CoreAdmin STATUS request in CoreAdminOperation#getCoreStatus. The instancePath is not retrieved from the ResourceLoader anymore, but from the registered CoreDescriptor. See commit [1]. SolrCore.getInstancePath(SolrCore.java:333) throws an NPE because the CoreContainer does not have a CoreDescriptor for the name, even though a SolrCore is available in the CoreContainer under that name (retrieved some lines above). This inconsistency is persistent: All STATUS requests keep failing until Solr is restarted. IIUC, the underlying problem is that CoreContainer#create does not correctly handle concurrent requests to create the same core. There's a race condition (see TODO comment [2]), and CoreContainer#createFromDescriptor may be called subsequently for the same core. The second call then fails to create an IndexWriter (LockObtainFailedException), and this causes a call to SolrCores#removeCoreDescriptor [3]. This mean, the second call removes the CoreDescriptor for the SolrCore created with the first call. This is the inconsistency that causes the NPE in CoreAdminOperation#getCoreStatus. Does this sound reasonable? I'll create a JIRA ticket tomorrow, if that's okay. Thank you, Andreas [1] https://github.com/apache/lucene-solr/commit/17ae79b0905b2bf8635c1b260b30807cae2f5463#diff-9652fe8353b7eff59cd6f128bb2699d88361e670b840ee5ca1018b1bc45584d1R324 [2] https://github.com/apache/lucene-solr/blob/15241573d3c8da0db3dfd380d99e4efcfe500c2e/solr/core/src/java/org/apache/solr/core/CoreContainer.java#L1242 [3] https://github.com/apache/lucene-solr/blob/15241573d3c8da0db3dfd380d99e4efcfe500c2e/solr/core/src/java/org/apache/solr/core/CoreContainer.java#L1407 -- Sent from: https://lucene.472066.n3.nabble.com/Solr-User-f472068.html
