murblanc commented on a change in pull request #1528:
URL: https://github.com/apache/lucene-solr/pull/1528#discussion_r430012046
##########
File path: solr/core/src/java/org/apache/solr/cloud/ZkController.java
##########
@@ -491,6 +494,41 @@ public boolean isClosed() {
assert ObjectReleaseTracker.track(this);
}
+ /**
+ * <p>Verifies if /clusterstate.json exists in Zookeepeer, and if it does
and is not empty, refuses to start and outputs
+ * a helpful message regarding collection migration.</p>
+ *
+ * <p>If /clusterstate.json exists and is empty, it is removed.</p>
+ */
+ private void checkNoOldClusterstate(final SolrZkClient zkClient) throws
InterruptedException {
+ try {
+ if (!zkClient.exists(ZkStateReader.UNSUPPORTED_CLUSTER_STATE, true)) {
+ return;
+ }
+
+ final byte[] data =
zkClient.getData(ZkStateReader.UNSUPPORTED_CLUSTER_STATE, null, null, true);
+
+ if (Arrays.equals("{}".getBytes(StandardCharsets.UTF_8), data)) {
+ // Empty json. This log will only occur once.
+ log.warn("{} no longer supported starting with Solr 9. Found empty
file on Zookeeper, deleting it.", ZkStateReader.UNSUPPORTED_CLUSTER_STATE);
+ zkClient.delete(ZkStateReader.UNSUPPORTED_CLUSTER_STATE, -1, true);
Review comment:
Looking at pre PR master branch.
The watcher on /clusterstate.json is an instance of
LegacyClusterStateWatcher (subclass of ZkStateReader).
The watcher processing is done in refreshAndWatch() that calls
ZkStateReader.refreshLegacyClusterState() and does some exception handling.
Even though refreshAndWatch() handles KeeperException.NoNodeException by
throwing a SolrException SERVICE_UNAVAILABLE, this never happens:
refreshLegacyClusterState() catches that exception, a comment says "Ignore
missing legacy clusterstate.json." and the catch builds what would be an empty
clusterstate.
We should be fine.
##########
File path: solr/solrj/src/java/org/apache/solr/common/cloud/ClusterState.java
##########
@@ -210,47 +200,42 @@ public boolean liveNodesContain(String name) {
@Override
public String toString() {
StringBuilder sb = new StringBuilder();
- sb.append("znodeVersion: ").append(znodeVersion);
- sb.append("\n");
sb.append("live nodes:").append(liveNodes);
sb.append("\n");
sb.append("collections:").append(collectionStates);
return sb.toString();
}
- public static ClusterState load(Integer version, byte[] bytes, Set<String>
liveNodes) {
- return load(version, bytes, liveNodes, ZkStateReader.CLUSTER_STATE);
- }
/**
- * Create ClusterState from json string that is typically stored in
zookeeper.
+ * Create a ClusterState from Json.
*
- * @param version zk version of the clusterstate.json file (bytes)
- * @param bytes clusterstate.json as a byte array
+ * @param bytes a byte array of a Json representation of a mapping from
collection name to the Json representation of a
+ * {@link DocCollection} as written by {@link
#write(JSONWriter)}. It can represent
+ * one or more collections.
* @param liveNodes list of live nodes
* @return the ClusterState
*/
- public static ClusterState load(Integer version, byte[] bytes, Set<String>
liveNodes, String znode) {
- // System.out.println("######## ClusterState.load:" + (bytes==null ? null
: new String(bytes)));
+ public static ClusterState createFromJson(int version, byte[] bytes,
Set<String> liveNodes) {
if (bytes == null || bytes.length == 0) {
- return new ClusterState(version, liveNodes, Collections.<String,
DocCollection>emptyMap());
+ return new ClusterState(liveNodes, Collections.<String,
DocCollection>emptyMap());
}
Map<String, Object> stateMap = (Map<String, Object>) Utils.fromJSON(bytes);
- return load(version, stateMap, liveNodes, znode);
+ return createFromData(version, stateMap, liveNodes);
}
- public static ClusterState load(Integer version, Map<String, Object>
stateMap, Set<String> liveNodes, String znode) {
+ public static ClusterState createFromData(int version, Map<String, Object>
stateMap, Set<String> liveNodes) {
Review comment:
`createFromCollectionMap`
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
