Apache Solr open-source search software
SOLR-15157: fix wrong assumptions on stats returned by Overseer when cluster state updates are distributed (#2410)
Ilan Ginzburgc472be5b8687
diff --git a/solr/core/src/java/org/apache/solr/cloud/api/collections/OverseerStatusCmd.java b/solr/core/src/java/org/apache/solr/cloud/api/collections/OverseerStatusCmd.javaindex 4b3dcc28b7a..c7efe4b8ae3 100644--- a/solr/core/src/java/org/apache/solr/cloud/api/collections/OverseerStatusCmd.java+++ b/solr/core/src/java/org/apache/solr/cloud/api/collections/OverseerStatusCmd.java@@ -36,6 +36,96 @@ import org.apache.zookeeper.data.Stat;import org.slf4j.Logger;import org.slf4j.LoggerFactory;+/**+ * This command returns stats about the Overseer, the cluster state updater and collection API activity occurring+ * <b>within the current Overseer node</b> (this is important because distributed operations occurring on other nodes+ * are <b>not included</b> in these stats, for example distributed cluster state updates or Per Replica States updates).<p>+ *+ * The structure of the returned results is as follows:+ * <ul>+ * <li><b>{@code leader}:</b> {@code ID} of the current overseer leader node</li>+ * <li><b>{@code overseer_queue_size}:</b> count of entries in the {@code /overseer/queue} Zookeeper queue/directory</li>+ * <li><b>{@code overseer_work_queue_size}:</b> count of entries in the {@code /overseer/queue-work} Zookeeper queue/directory</li>+ * <li><b>{@code overseer_collection_queue_size}:</b> count of entries in the {@code /overseer/collection-queue-work}+ * Zookeeper queue/directory</li>+ * <li><b>{@code overseer_operations}:</b> map (of maps) of success and error counts for operations. The operations+ * (keys) tracked in this map are:+ * <ul>+ * <li>{@code am_i_leader} (Overseer checking it is still the elected Overseer as it processes cluster state update+ * messages)</li>+ * <li>{@code configset_}<i>{@code <config set operation>}</i> (from+ * {@link org.apache.solr.handler.admin.ConfigSetsHandler.ConfigSetOperation})</li>+ * <li>Cluster state change operation names from {@link org.apache.solr.common.params.CollectionParams.CollectionAction}+ * (not all of them!) and {@link org.apache.solr.cloud.overseer.OverseerAction} (the complete list: {@code create},+ * {@code delete}, {@code createshard}, {@code deleteshard}, {@code addreplica}, {@code addreplicaprop}, {@code deletereplicaprop},+ * {@code balanceshardunique}, {@code modifycollection}, {@code state}, {@code leader}, {@code deletecore}, {@code addroutingrule},+ * {@code removeroutingrule}, {@code updateshardstate}, {@code downnode} and {@code quit} with this like one unlikely+ * to be observed since the Overseer is existing right away)</li>+ * <li>{@code update_state} (when Overseer cluster state updater persists changes in Zookeeper)</li>+ * </ul>+ * For each key, the value is a map composed of:+ * <ul>+ * <li>{@code requests}: success count of the given operation </li>+ * <li>{@code errors}: error count of the operation </li>+ * <li>More metrics (see below)</li>+ * </ul>+ * </li>+ * <li><b>{@code collection_operations}:</b> map (of maps) of success and error counts for collection related operations.+ * The operations(keys) tracked in this map are <b>all operations</b> that start with {@code collection_}, but the+ * {@code collection_} prefix is <b>stripped</b> of the returned value. Possible keys are therefore:+ * <ul>+ * <li>{@code am_i_leader}: originating in a stat called {@code collection_am_i_leader} representing Overseer checking+ * it is still the elected Overseer as it processes Collection API and Config Set API messages.</li>+ * <li>Collection API operation names from {@link org.apache.solr.common.params.CollectionParams.CollectionAction} (the+ * stripped {@code collection_} prefix gets added in {@link OverseerCollectionMessageHandler#getTimerName(String)})</li>+ * </ul>+ * For each key, the value is a map composed of:+ * <ul>+ * <li>{@code requests}: success count of the given operation </li>+ * <li>{@code errors}: error count of the operation </li>+ * <li>{@code recent_failures}: an <b>optional</b> entry containing a list of maps, each map having two entries, one+ * with key {@code request} with a failed request properties (a {@link ZkNodeProps}) and the other with key+ * {@code response} with the corresponding response properties (a {@link org.apache.solr.client.solrj.SolrResponse}).</li>+ * <li>More metrics (see below)</li>+ * </ul>+ * </li>+ * <li><b>{@code overseer_queue}:</b> metrics on operations done on the Zookeeper queue {@code /overseer/queue} (see+ * metrics below).<br>+ * The operations that can be done on the queue and that can be keys whose values are a metrics map are:+ * <ul>+ * <li>{@code offer}</li>+ * <li>{@code peek}</li>+ * <li>{@code peek_wait}</li>+ * <li>{@code peek_wait_forever}</li>+ * <li>{@code peekTopN_wait}</li>+ * <li>{@code peekTopN_wait_forever}</li>+ * <li>{@code poll}</li>+ * <li>{@code remove}</li>+ * <li>{@code remove_event}</li>+ * <li>{@code take}</li>+ * </ul>+ * </li>+ * <li><b>{@code overseer_internal_queue}:</b> same as above but for queue {@code /overseer/queue-work}</li>+ * <li><b>{@code collection_queue}:</b> same as above but for queue {@code /overseer/collection-queue-work}</li>+ * </ul>+ *+ * <p>+ * Maps returned as values of keys in <b>{@code overseer_operations}</b>, <b>{@code collection_operations}</b>,+ * <b>{@code overseer_queue}</b>, <b>{@code overseer_internal_queue}</b> and <b>{@code collection_queue}</b> include+ * additional stats. These stats are provided by {@link MetricUtils}, and represent metrics on each type of operation+ * execution (be it failed or successful), see calls to {@link Stats#time(String)}. The metric keys are:+ * <ul>+ * <li>{@code avgRequestsPerSecond}</li>+ * <li>{@code 5minRateRequestsPerSecond}</li>+ * <li>{@code 15minRateRequestsPerSecond}</li>+ * <li>{@code avgTimePerRequest}</li>+ * <li>{@code medianRequestTime}</li>+ * <li>{@code 75thPcRequestTime}</li>+ * <li>{@code 95thPcRequestTime}</li>+ * <li>{@code 99thPcRequestTime}</li>+ * <li>{@code 999thPcRequestTime}</li>+ * </ul>+ */public class OverseerStatusCmd implements CollApiCmds.CollectionApiCommand {private static final Logger log = LoggerFactory.getLogger(MethodHandles.lookup().lookupClass());private final CollectionCommandContext ccc;@@ -60,16 +150,6 @@ public class OverseerStatusCmd implements CollApiCmds.CollectionApiCommand {zkStateReader.getZkClient().getData("/overseer/collection-queue-work",null, stat, true);results.add("overseer_collection_queue_size", stat.getNumChildren());- // Overseer reported stats below are tracked in the Overseer cluster state updater when it performs certain operations.- // Sharing the ocmh.stats variable between the cluster state updater and the Collection API (this command) is by the way- // about the only thing that ties the cluster state updater to the collection api message handler and that takes- // advantage of the fact that both run on the same node (the Overseer node). (recently added PerReplicaStates also- // take advantage of this through method Overseer.submit() accessed via CollectionCommandContext.submitIntraProcessMessage()).- // When distributed updates are enabled, cluster state updates are not done by the Overseer (it doesn't even see them)- // and therefore can't report them. The corresponding data in OVERSEERSTATUS (all data built below) is no longer returned.- // This means top level keys "overseer_operations", "collection_operations", "overseer_queue", "overseer_internal_queue"- // and "collection_queue" are either empty or do not contain all expected information when cluster state updates are distributed.-@SuppressWarnings({"rawtypes"})NamedList overseerStats = new NamedList();@SuppressWarnings({"rawtypes"})@@ -80,47 +160,45 @@ public class OverseerStatusCmd implements CollApiCmds.CollectionApiCommand {NamedList workQueueStats = new NamedList();@SuppressWarnings({"rawtypes"})NamedList collectionQueueStats = new NamedList();- // stats below do not make sense when cluster state updates are distributed. Return them empty.- if (!ccc.getDistributedClusterStateUpdater().isDistributedStateUpdate()) {- Stats stats = ccc.getOverseerStats();- for (Map.Entry<String, Stats.Stat> entry : stats.getStats().entrySet()) {- String key = entry.getKey();- NamedList<Object> lst = new SimpleOrderedMap<>();- if (key.startsWith("collection_")) {- collectionStats.add(key.substring(11), lst);- int successes = stats.getSuccessCount(entry.getKey());- int errors = stats.getErrorCount(entry.getKey());- lst.add("requests", successes);- lst.add("errors", errors);- List<Stats.FailedOp> failureDetails = stats.getFailureDetails(key);- if (failureDetails != null) {- List<SimpleOrderedMap<Object>> failures = new ArrayList<>();- for (Stats.FailedOp failedOp : failureDetails) {- SimpleOrderedMap<Object> fail = new SimpleOrderedMap<>();- fail.add("request", failedOp.req.getProperties());- fail.add("response", failedOp.resp.getResponse());- failures.add(fail);- }- lst.add("recent_failures", failures);+ Stats stats = ccc.getOverseerStats();+ for (Map.Entry<String, Stats.Stat> entry : stats.getStats().entrySet()) {+ String key = entry.getKey();+ NamedList<Object> lst = new SimpleOrderedMap<>();+ if (key.startsWith("collection_")) {+ collectionStats.add(key.substring(11), lst);+ int successes = stats.getSuccessCount(entry.getKey());+ int errors = stats.getErrorCount(entry.getKey());+ lst.add("requests", successes);+ lst.add("errors", errors);+ List<Stats.FailedOp> failureDetails = stats.getFailureDetails(key);+ if (failureDetails != null) {+ List<SimpleOrderedMap<Object>> failures = new ArrayList<>();+ for (Stats.FailedOp failedOp : failureDetails) {+ SimpleOrderedMap<Object> fail = new SimpleOrderedMap<>();+ fail.add("request", failedOp.req.getProperties());+ fail.add("response", failedOp.resp.getResponse());+ failures.add(fail);}- } else if (key.startsWith("/overseer/queue_")) {- stateUpdateQueueStats.add(key.substring(16), lst);- } else if (key.startsWith("/overseer/queue-work_")) {- workQueueStats.add(key.substring(21), lst);- } else if (key.startsWith("/overseer/collection-queue-work_")) {- collectionQueueStats.add(key.substring(32), lst);- } else {- // overseer stats- overseerStats.add(key, lst);- int successes = stats.getSuccessCount(entry.getKey());- int errors = stats.getErrorCount(entry.getKey());- lst.add("requests", successes);- lst.add("errors", errors);+ lst.add("recent_failures", failures);}- Timer timer = entry.getValue().requestTime;- MetricUtils.addMetrics(lst, timer);+ } else if (key.startsWith("/overseer/queue_")) {+ stateUpdateQueueStats.add(key.substring(16), lst);+ } else if (key.startsWith("/overseer/queue-work_")) {+ workQueueStats.add(key.substring(21), lst);+ } else if (key.startsWith("/overseer/collection-queue-work_")) {+ collectionQueueStats.add(key.substring(32), lst);+ } else {+ // overseer stats+ overseerStats.add(key, lst);+ int successes = stats.getSuccessCount(entry.getKey());+ int errors = stats.getErrorCount(entry.getKey());+ lst.add("requests", successes);+ lst.add("errors", errors);}+ Timer timer = entry.getValue().requestTime;+ MetricUtils.addMetrics(lst, timer);}+results.add("overseer_operations", overseerStats);results.add("collection_operations", collectionStats);results.add("overseer_queue", stateUpdateQueueStats);diff --git a/solr/core/src/test/org/apache/solr/cloud/OverseerStatusTest.java b/solr/core/src/test/org/apache/solr/cloud/OverseerStatusTest.javaindex 23402544e8c..d2d96047658 100644--- a/solr/core/src/test/org/apache/solr/cloud/OverseerStatusTest.java+++ b/solr/core/src/test/org/apache/solr/cloud/OverseerStatusTest.java@@ -48,9 +48,11 @@ public class OverseerStatusTest extends SolrCloudTestCase {NamedList<Object> overseer_operations = (NamedList<Object>) resp.get("overseer_operations");SimpleOrderedMap<Object> createcollection= (SimpleOrderedMap<Object>) collection_operations.get(CollectionParams.CollectionAction.CREATE.toLower());- // When cluster state updates are distributed, Overseer doesn't see the updates and doesn't report stats on them.+ assertEquals("No stats for create in OverseerCollectionProcessor", numCollectionCreates + 1, createcollection.get("requests"));+ // When cluster state updates are distributed, Overseer doesn't see them and doesn't report stats on them.if (!cluster.getOpenOverseer().getDistributedClusterStateUpdater().isDistributedStateUpdate()) {- assertEquals("No stats for create in OverseerCollectionProcessor", numCollectionCreates + 1, createcollection.get("requests"));+ // Note the "create" key here is in a different map from the "create" key above. Above it's Collection creation in the+ // Collection API, here it's the collection creation from the cluster state updater perspective.createcollection = (SimpleOrderedMap<Object>) overseer_operations.get(CollectionParams.CollectionAction.CREATE.toLower());assertEquals("No stats for create in Overseer", numOverseerCreates + 1, createcollection.get("requests"));}