Merge pull request #42 from MarginaliaSearch/no-downtime-upgrades

Zero downtime upgrades, merge-based index construction

build.gradle

@@ -13,6 +13,10 @@ tasks.register('dist', Copy) {
     from subprojects.collect { it.tasks.withType(Tar) }
     into "$buildDir/dist"
 
+    // For local development, each processes that are to be triggerable
+    // from the control-service need to go here to end up somewhere the
+    // control-service can find them
+
     doLast {
         copy {
             from tarTree("$buildDir/dist/converter-process.tar")
@@ -34,10 +38,18 @@ tasks.register('dist', Copy) {
             from tarTree("$buildDir/dist/crawl-job-extractor-process.tar")
             into "$projectDir/run/dist/"
         }
+        copy {
+            from tarTree("$buildDir/dist/index-construction-process.tar")
+            into "$projectDir/run/dist/"
+        }
     }
 }
 idea {
     module {
+        // Exclude these directories from being indexed by IntelliJ
+        // as they tend to bring the IDE to its knees and use up all
+        // Inotify spots in a hurry
+        excludeDirs.add(file("$projectDir/run/backup"))
         excludeDirs.add(file("$projectDir/run/model"))
         excludeDirs.add(file("$projectDir/run/dist"))
         excludeDirs.add(file("$projectDir/run/samples"))

code/api/index-api/src/main/java/nu/marginalia/index/client/IndexMqEndpoints.java

@@ -3,8 +3,6 @@
 public class IndexMqEndpoints {
     public static final String INDEX_IS_BLOCKED = "INDEX-IS-BLOCKED";
     public static final String INDEX_REPARTITION = "INDEX-REPARTITION";
-
-    public static final String INDEX_RELOAD_LEXICON = "INDEX-RELOAD-LEXICON";
-    public static final String INDEX_REINDEX = "INDEX-REINDEX";
+    public static final String SWITCH_INDEX = "SWITCH-INDEX";
 
 }

code/api/index-api/src/main/java/nu/marginalia/index/client/model/results/SearchResultItem.java

@@ -2,16 +2,17 @@
 
 import lombok.AllArgsConstructor;
 import lombok.Getter;
-import nu.marginalia.model.EdgeUrl;
-import nu.marginalia.model.id.EdgeId;
+import nu.marginalia.model.id.UrlIdCodec;
+import org.jetbrains.annotations.NotNull;
 
 import java.util.ArrayList;
 import java.util.List;
 
 /** Represents a document matching a search query */
 @AllArgsConstructor @Getter
-public class SearchResultItem {
-    /** Encoded ID that contains both the URL id and its ranking */
+public class SearchResultItem implements Comparable<SearchResultItem> {
+    /** Encoded ID that contains both the URL id and its ranking.  This is
+     * probably not what you want, use getDocumentId() instead */
     public final long combinedId;
 
     /** How did the subqueries match against the document ? */
@@ -20,20 +21,18 @@ public class SearchResultItem {
     /** How many other potential results existed in the same domain */
     public int resultsFromDomain;
 
-    public SearchResultItem(long val) {
-        this.combinedId = val;
+    public SearchResultItem(long combinedId) {
+        this.combinedId = combinedId;
         this.keywordScores = new ArrayList<>(16);
     }
 
-    public EdgeId<EdgeUrl> getUrlId() {
-        return new EdgeId<>(getUrlIdInt());
-    }
 
-    public int getUrlIdInt() {
-        return (int)(combinedId & 0xFFFF_FFFFL);
+    public long getDocumentId() {
+        return UrlIdCodec.removeRank(combinedId);
     }
+
     public int getRanking() {
-        return (int)(combinedId >>> 32);
+        return UrlIdCodec.getRank(combinedId);
     }
 
     /* Used for evaluation */
@@ -45,20 +44,16 @@ public SearchResultPreliminaryScore getScore() {
         return scoreValue;
     }
 
-    private transient int domainId = Integer.MIN_VALUE;
-    public void setDomainId(int domainId) {
-        this.domainId = domainId;
-    }
     public int getDomainId() {
-        return this.domainId;
+        return UrlIdCodec.getDomainId(this.combinedId);
     }
 
     public int hashCode() {
-        return getUrlIdInt();
+        return Long.hashCode(combinedId);
     }
 
     public String toString() {
-        return getClass().getSimpleName() + "[ url= " + getUrlId() + ", rank=" + getRanking() + "]";
+        return getClass().getSimpleName() + "[ url= " + getDocumentId() + ", rank=" + getRanking() + "]";
     }
 
     public boolean equals(Object other) {
@@ -67,18 +62,18 @@ public boolean equals(Object other) {
         if (other == this)
             return true;
         if (other instanceof SearchResultItem o) {
-            return o.getUrlIdInt()  == getUrlIdInt();
+            return o.getDocumentId()  == getDocumentId();
         }
         return false;
     }
 
-    public long deduplicationKey() {
-        final int domainId = getDomainId();
-
-        if (domainId == Integer.MAX_VALUE || domainId == Integer.MIN_VALUE) {
-            return 0;
-        }
+    @Override
+    public int compareTo(@NotNull SearchResultItem o) {
+        // this looks like a bug, but we actually want this in a reversed order
+        int diff = o.getScore().compareTo(getScore());
+        if (diff != 0)
+            return diff;
 
-        return domainId;
+        return Long.compare(this.combinedId, o.combinedId);
     }
 }

code/api/index-api/src/main/java/nu/marginalia/index/client/model/results/SearchResultPreliminaryScore.java

@@ -6,7 +6,6 @@
 import static java.lang.Double.compare;
 
 public record SearchResultPreliminaryScore(
-        boolean disqualified,
         boolean hasPriorityTerm,
         double searchRankingScore)
         implements Comparable<SearchResultPreliminaryScore>
@@ -25,7 +24,4 @@ public int compareTo(@NotNull SearchResultPreliminaryScore other) {
         return PREFER_LOW * compare(searchRankingScore, other.searchRankingScore);
     }
 
-    public boolean isDisqualified() {
-        return disqualified;
-    }
 }

code/api/process-mqapi/src/main/java/nu/marginalia/mqapi/ProcessInboxNames.java

@@ -4,4 +4,6 @@ public class ProcessInboxNames {
     public static final String CONVERTER_INBOX = "converter";
     public static final String LOADER_INBOX = "loader";
     public static final String CRAWLER_INBOX = "crawler";
+
+    public static final String INDEX_CONSTRUCTOR_INBOX = "index_constructor";
 }

code/api/process-mqapi/src/main/java/nu/marginalia/mqapi/index/CreateIndexRequest.java

@@ -0,0 +1,5 @@
+package nu.marginalia.mqapi.index;
+
+public record CreateIndexRequest(IndexName indexName)
+{
+}

code/api/process-mqapi/src/main/java/nu/marginalia/mqapi/index/IndexName.java

@@ -0,0 +1,7 @@
+package nu.marginalia.mqapi.index;
+
+public enum IndexName {
+    FORWARD,
+    REVERSE_FULL,
+    REVERSE_PRIO
+}

code/api/search-api/src/main/java/nu/marginalia/search/client/SearchMqEndpoints.java

@@ -3,4 +3,5 @@
 public class SearchMqEndpoints {
     /** Flushes the URL caches, run if significant changes have occurred in the URLs database */
     public static final String FLUSH_CACHES = "FLUSH_CACHES";
+    public static final String SWITCH_LINKDB = "SWITCH_LINKDB";
 }

code/common/db/src/main/java/nu/marginalia/db/DbDomainQueries.java

@@ -9,16 +9,16 @@
 import com.zaxxer.hikari.HikariDataSource;
 import lombok.SneakyThrows;
 import nu.marginalia.model.EdgeDomain;
-import nu.marginalia.model.id.EdgeId;
 
 import java.util.NoSuchElementException;
 import java.util.Optional;
+import java.util.OptionalInt;
 
 @Singleton
 public class DbDomainQueries {
     private final HikariDataSource dataSource;
 
-    private final Cache<EdgeDomain, EdgeId<EdgeDomain>> domainIdCache = CacheBuilder.newBuilder().maximumSize(10_000).build();
+    private final Cache<EdgeDomain, Integer> domainIdCache = CacheBuilder.newBuilder().maximumSize(10_000).build();
 
     @Inject
     public DbDomainQueries(HikariDataSource dataSource)
@@ -28,15 +28,15 @@ public DbDomainQueries(HikariDataSource dataSource)
 
 
     @SneakyThrows
-    public EdgeId<EdgeDomain> getDomainId(EdgeDomain domain) {
+    public Integer getDomainId(EdgeDomain domain) {
         try (var connection = dataSource.getConnection()) {
 
             return domainIdCache.get(domain, () -> {
                 try (var stmt = connection.prepareStatement("SELECT ID FROM EC_DOMAIN WHERE DOMAIN_NAME=?")) {
                     stmt.setString(1, domain.toString());
                     var rsp = stmt.executeQuery();
                     if (rsp.next()) {
-                        return new EdgeId<>(rsp.getInt(1));
+                        return rsp.getInt(1);
                     }
                 }
                 throw new NoSuchElementException();
@@ -48,38 +48,38 @@ public EdgeId<EdgeDomain> getDomainId(EdgeDomain domain) {
     }
 
     @SneakyThrows
-    public Optional<EdgeId<EdgeDomain>> tryGetDomainId(EdgeDomain domain) {
+    public OptionalInt tryGetDomainId(EdgeDomain domain) {
 
-        var maybe = Optional.ofNullable(domainIdCache.getIfPresent(domain));
-
-        if (maybe.isPresent())
-            return maybe;
+        Integer maybeId = domainIdCache.getIfPresent(domain);
+        if (maybeId != null) {
+            return OptionalInt.of(maybeId);
+        }
 
         try (var connection = dataSource.getConnection()) {
 
             try (var stmt = connection.prepareStatement("SELECT ID FROM EC_DOMAIN WHERE DOMAIN_NAME=?")) {
                 stmt.setString(1, domain.toString());
                 var rsp = stmt.executeQuery();
                 if (rsp.next()) {
-                    var id = new EdgeId<EdgeDomain>(rsp.getInt(1));
+                    var id = rsp.getInt(1);
 
                     domainIdCache.put(domain, id);
-                    return Optional.of(id);
+                    return OptionalInt.of(id);
                 }
             }
-            return Optional.empty();
+            return OptionalInt.empty();
         }
         catch (UncheckedExecutionException ex) {
-            return Optional.empty();
+            return OptionalInt.empty();
         }
     }
 
     @SneakyThrows
-    public Optional<EdgeDomain> getDomain(EdgeId<EdgeDomain> id) {
+    public Optional<EdgeDomain> getDomain(int id) {
         try (var connection = dataSource.getConnection()) {
 
             try (var stmt = connection.prepareStatement("SELECT DOMAIN_NAME FROM EC_DOMAIN WHERE ID=?")) {
-                stmt.setInt(1, id.id());
+                stmt.setInt(1, id);
                 var rsp = stmt.executeQuery();
                 if (rsp.next()) {
                     return Optional.of(new EdgeDomain(rsp.getString(1)));

code/common/db/src/main/java/nu/marginalia/db/DomainBlacklist.java

@@ -2,15 +2,10 @@
 
 import com.google.inject.ImplementedBy;
 import gnu.trove.set.hash.TIntHashSet;
-import nu.marginalia.model.EdgeDomain;
-import nu.marginalia.model.id.EdgeId;
 
 @ImplementedBy(DomainBlacklistImpl.class)
 public interface DomainBlacklist {
     boolean isBlacklisted(int domainId);
-    default boolean isBlacklisted(EdgeId<EdgeDomain> domainId) {
-        return isBlacklisted(domainId.id());
-    }
     default TIntHashSet getSpamDomains() {
         return new TIntHashSet();
     }

code/common/db/src/main/java/nu/marginalia/db/DomainTypes.java

@@ -1,13 +1,14 @@
 package nu.marginalia.db;
 
 import com.zaxxer.hikari.HikariDataSource;
+import gnu.trove.list.TIntList;
+import gnu.trove.list.array.TIntArrayList;
 import nu.marginalia.model.EdgeDomain;
-import nu.marginalia.model.id.EdgeIdList;
 import org.slf4j.LoggerFactory;
 import org.slf4j.Logger;
 
-import javax.inject.Inject;
-import javax.inject.Singleton;
+import com.google.inject.Inject;
+import com.google.inject.Singleton;
 import java.io.BufferedReader;
 import java.io.IOException;
 import java.io.InputStreamReader;
@@ -58,10 +59,10 @@ public List<String> getAllDomainsByType(Type type) {
         return ret;
     }
 
-    /** Retrieve the EdgeId of all domains of a certain type,
+    /** Retrieve the domain id of all domains of a certain type,
      * ignoring entries that are not in the EC_DOMAIN table */
-    public EdgeIdList<EdgeDomain> getKnownDomainsByType(Type type) {
-        EdgeIdList<EdgeDomain> ret = new EdgeIdList<>();
+    public TIntList getKnownDomainsByType(Type type) {
+        TIntList ret = new TIntArrayList();
 
         try (var conn = dataSource.getConnection();
              var stmt = conn.prepareStatement("""

code/common/db/src/main/java/nu/marginalia/db/storage/FileStorageService.java

@@ -5,8 +5,8 @@
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
-import javax.inject.Inject;
-import javax.inject.Singleton;
+import com.google.inject.Inject;
+import com.google.inject.Singleton;
 import java.io.File;
 import java.io.FileNotFoundException;
 import java.io.IOException;

code/common/db/src/main/java/nu/marginalia/db/storage/model/FileStorageType.java

@@ -5,9 +5,9 @@ public enum FileStorageType {
     CRAWL_DATA,
     PROCESSED_DATA,
     INDEX_STAGING,
-    LEXICON_STAGING,
+    LINKDB_STAGING,
+    LINKDB_LIVE,
     INDEX_LIVE,
-    LEXICON_LIVE,
     BACKUP,
     EXPORT,
     SEARCH_SETS

code/common/db/src/main/resources/db/migration/V23_09_0_000__filestorage_livedb.sql

@@ -0,0 +1,9 @@
+ALTER TABLE FILE_STORAGE MODIFY COLUMN TYPE ENUM ('CRAWL_SPEC', 'CRAWL_DATA', 'PROCESSED_DATA', 'INDEX_STAGING', 'LEXICON_STAGING', 'INDEX_LIVE', 'LEXICON_LIVE', 'SEARCH_SETS', 'BACKUP', 'EXPORT', 'LINKDB_LIVE', 'LINKDB_STAGING') NOT NULL;
+
+INSERT IGNORE INTO FILE_STORAGE(BASE_ID, PATH, DESCRIPTION, TYPE)
+SELECT ID, 'ldbr', "Linkdb Current", 'LINKDB_LIVE'
+FROM FILE_STORAGE_BASE WHERE NAME='Index Storage';
+
+INSERT IGNORE INTO FILE_STORAGE(BASE_ID, PATH, DESCRIPTION, TYPE)
+SELECT ID, 'ldbw', "Linkdb Staging Area", 'LINKDB_STAGING'
+FROM FILE_STORAGE_BASE WHERE NAME='Index Storage';

code/common/db/src/main/resources/db/migration/V23_09_1_000__drop_ecurl.sql

@@ -0,0 +1,3 @@
+DROP VIEW EC_URL_VIEW;
+DROP TABLE EC_PAGE_DATA;
+DROP TABLE EC_URL;

code/common/db/src/main/resources/db/migration/V23_09_2_000__filestorage_backup.sql

@@ -0,0 +1,3 @@
+INSERT IGNORE INTO FILE_STORAGE_BASE(NAME, PATH, TYPE, PERMIT_TEMP)
+VALUES
+('Backup Storage', '/backup', 'BACKUP', true);

code/common/db/src/main/resources/db/migration/V23_09_2_001__filestorage_no_lexicon.sql

@@ -0,0 +1 @@
+DELETE FROM FILE_STORAGE WHERE TYPE IN ('LEXICON_STAGING', 'LEXICON_LIVE');