5 Replies Latest reply on Jun 19, 2018 6:13 AM by gustavonalle

    Strange problem of distributed jdbc cache store for index.

    seto

      private ConfigurationBuilder persistenceConfigLucene(ConfigurationBuilder configurationBuilder) {

        configurationBuilder.persistence()

        .addStore(JdbcStringBasedStoreConfigurationBuilder.class)

        .key2StringMapper(LuceneKey2StringMapper.class)

        .async().enable()

        .preload(preload)

        .shared(shared)

        .fetchPersistentState(fetchPersistentState)

        .ignoreModifications(ignoreModifications)

        .purgeOnStartup(purgeOnStartup)

        .table()

        .createOnStart(createOnStart)

        .dropOnExit(dropOnExit)

        .tableNamePrefix("ISPN_STRING_TABLE")

        .idColumnName("ID_COLUMN").idColumnType("VARCHAR(255)")

        .dataColumnName("DATA_COLUMN").dataColumnType("BLOB")

        .timestampColumnName("TIMESTAMP_COLUMN").timestampColumnType("BIGINT")

        .connectionPool()

        .connectionUrl(connectionUrl)

        .username(username)

        .password(password)

        .driverClass(driverClass);
        return configurationBuilder;
      }

      preload = true, shared = true, fetchPersistentState = true, ignoreModifications = false, purgeOnStartup = true

       

      It seems that I will got incorrect remaining data even purgeOnStartup is true. PurgeOnStartup for debug purpose.

       

      public void sendStoryList(DataRef<User> userDataRef, DataRef<ClientChannel> clientChannelDataRef) {

        SearchManager searchManager = distributedDataAPI.getSearchManager();
         QueryBuilder queryBuilder = searchManager.buildQueryBuilderForClass(StoryPending.class).get();
         Query query = queryBuilder.keyword().onField("user").matching(userDataRef.get().getId()).createQuery();
         System.out.println("sendList:" + searchManager.getQuery(query, StoryPending.class).getResultSize());
         CacheQuery<StoryPending> cacheQuery = searchManager.getQuery(query, StoryPending.class);
         List<StoryPending> storyPendings = cacheQuery.list();
         System.out.println("sendList:" + storyPendings.size());
         DimensionStory.StoryList.Builder builder = DimensionStory.StoryList.newBuilder();
        for (StoryPending storyPending : storyPendings) {

        System.out.println("sendList:" + storyPending.getStory());
         DimensionStory.Story.Builder builder0 = DimensionStory.Story.newBuilder().setId(storyPending.getStory());
         builder.addStories(builder0);
         }

        clientChannelDataRef.get().send(builder);
      }

      The getResultSize will get 1, but the second is 0.

      I thought that there should be not remaining incorrect data if purgeOnStartue is true.

       

      No remaining incorrect data if I deleted the tables in the mysql database before the second launch.

      Yet, my project is complicated. So I can't provide a reproducing project.

       

      But I'd like you guys to check this. Or do I misunderstand the purgeOnStartup in this case?

        • 1. Re: Strange problem of distributed jdbc cache store for index.
          galder.zamarreno

          What Infinispan version? Did you try enabling TRACE logging for org.infinispan package?

          • 2. Re: Strange problem of distributed jdbc cache store for index.
            seto

            9.2.4.Final.

            I didn't. How can I enabling TRACE only for org.infinispan?

            Here's my log4j2.xml. I don't know how to enable trace for org.infinispan only.

            If I change the root level to TRACE. A lot of other logging is printed.

            And also I'm using distributed executor service in loop. So maybe it would print unrelated info as well.

            I thought that the purge fail should be at least a warn.

            <?xml version="1.0" encoding="UTF-8"?>

            <Configuration status="INFO" packages="co.kaiba.blueeyes.impl.util.log4j">

                <Appenders>

                    <Console name="A1" target="SYSTEM_OUT">

                        <Filters>

                            <ThresholdFilter level="warn" onMatch="ACCEPT" onMismatch="NEUTRAL"/>

                            <!--<ClassNameRegexFilter regex="co\..*" onMatch="ACCEPT" onMismatch="DENY"/>-->

                        </Filters>

                        <PatternLayout pattern="%d [%t] %p %c - %m%n"/>

                    </Console>

                </Appenders>

                <Loggers>

                    <Root level="INFO">

                        <AppenderRef ref="A1"/>

                    </Root>

                </Loggers>

            </Configuration>

            • 4. Re: Strange problem of distributed jdbc cache store for index.
              seto

              Here's the log.

              Executing Lucene query 'user:78f3ebf0-33a4-47c9-9db1-b4b27e6887db'

              This query for StoryPending should be the one with problem.

               

              This is the lucene cache containers config.

              Can you see any problem?

              @ConfigureCache("LuceneIndexesMetadata")

              @Produces
              public Configuration luceneIndexesMetadataConfig() {

                ConfigurationBuilder configurationBuilder = new ConfigurationBuilder();
                 distributeConfig(persistenceConfigLucene(configurationBuilder));
                return configurationBuilder.build();
              }

               

              @ConfigureCache("LuceneIndexesData")

              @Produces
              public Configuration luceneIndexesDataConfig() {

                ConfigurationBuilder configurationBuilder = new ConfigurationBuilder();
                 distributeConfig(persistenceConfigLucene(configurationBuilder));
                return configurationBuilder.build();
              }

               

              @ConfigureCache("LuceneIndexesLocking")

              @Produces
              public Configuration luceneIndexesLockingConfig() {

                ConfigurationBuilder configurationBuilder = new ConfigurationBuilder();
                 distributeConfig(persistenceConfigLucene(configurationBuilder));
                return configurationBuilder.build();
              }

               

              private ConfigurationBuilder distributeConfig(ConfigurationBuilder builder) {

                builder

                .clustering()

                .cacheMode(CacheMode.DIST_SYNC);
                return builder;
              }

              private ConfigurationBuilder persistenceConfigLucene(ConfigurationBuilder configurationBuilder) {

                configurationBuilder.persistence()

                .addStore(JdbcStringBasedStoreConfigurationBuilder.class)

                .key2StringMapper(LuceneKey2StringMapper.class)

                .async().enable()

                .preload(preload)

                .shared(shared)

                .fetchPersistentState(fetchPersistentState)

                .ignoreModifications(ignoreModifications)

                .purgeOnStartup(purgeOnStartup)

                .table()

                .createOnStart(createOnStart)

                .dropOnExit(dropOnExit)

                .tableNamePrefix("ISPN_STRING_TABLE")

                .idColumnName("ID_COLUMN").idColumnType("VARCHAR(255)")

                .dataColumnName("DATA_COLUMN").dataColumnType("BLOB")

                .timestampColumnName("TIMESTAMP_COLUMN").timestampColumnType("BIGINT")

                .connectionPool()

                .connectionUrl(connectionUrl)

                .username(username)

                .password(password)

                .driverClass(driverClass);
                return configurationBuilder;
              }

              • 5. Re: Strange problem of distributed jdbc cache store for index.
                gustavonalle

                Sorry, I did not understand what the problem is. What do you mean by "incorrect remaining data" ?

                Also, are you running Lucene queries using the distributed executor? This should not be necessary, as Infinispan has mechanism of broadcasting the query.

                 

                Could you send the full configuration of your indexed cache (not only the Lucene caches) and a step-by-step way to reproduce it? Example: start caches, write data, stop, query things like that.