PHOENIX-5211 Consistent Immutable Global Indexes for Non-Transactiona… #517

gokceni · 2019-06-11T23:14:09Z

…l Tables

swaroopak · 2019-06-12T05:09:13Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+            long mutationSizeBytes = 0;
+            long mutationCommitTime = 0;
+            long numFailedMutations = 0;
+            ;


nit: unnecessary ;

swaroopak · 2019-06-12T05:09:28Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+        }
+    }
+
+    private void sendMutations(Iterator<Entry<TableInfo, List<Mutation>>> mutationsIterator, Span span, ImmutableBytesWritable indexMetaDataPtr)


nice refactoring!

swaroopak · 2019-06-12T05:13:27Z

phoenix-core/src/main/java/org/apache/phoenix/util/IndexUtil.java

+        try {
+            desc = connection.getQueryServices().getTableDescriptor(index.getBytes());
+        } catch (SQLException ex) {
+            //logger.warn("Error during getTableDescriptor ", ex);


nit: commented code.

swaroopak · 2019-06-12T05:31:17Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+            }
+
+        }
+        return verifiedMutationsMap;


If I understood this function correctly, I find the variable name little misleading. verifiedMutationMap --> mutations in the map are verified, which doesn't seem to be the case. Rather, the mutations are from the first phase where verified column is set to false. Please correct me if I am wrong.

I see why it is confusing, I will do some more refactoring here

swaroopak · 2019-06-12T05:34:41Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+            PTable indexTable = null;
+            PTable dataTable = tableInfo.origTableRef.getTable();
+
+            List<PTable> indexes = dataTable.getIndexes();


This function will return all the indexes, including local indexes. Is that the expectation?

That is fine. It will not match the GlobalIndexChecker index name

swaroopak · 2019-06-12T05:42:11Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+
+            List<PTable> indexes = dataTable.getIndexes();
+            for (PTable index : indexes) {
+                if (index.getName().equals(tableInfo.hTableName)) {


Although private member hTableName is accessible in the Outer class, better to use the public method to access it.

priyankporwal · 2019-06-11T23:27:54Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+        upsertRows(conn, tableName, numOfRowsToInsert);
+        conn.commit();
+
+        Thread.sleep(15000);


Can we please avoid adding this sleep? I worry about increased test run-time every time we add such sleeps. Perhaps we can check readiness of the index/table - and/or query syscat, or other alternatives.

priyankporwal · 2019-06-11T23:28:32Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+                        + " ON " + tableName + " (long_pk, varchar_pk)"
+                        + " INCLUDE (long_col1, long_col2) ";
+
+        Statement stmt = conn.createStatement();


Is there a difference between the two createStatement() calls? If not, consistent usage would be nice.

priyankporwal · 2019-06-11T23:33:46Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+    private void createTableAndIndexForConsistentIndex(Connection conn, String tableName, String indexName, int numOfRowsToInsert)
+            throws Exception {
+        String ddl = "CREATE TABLE " + TABLE_NAME + TestUtil.TEST_TABLE_SCHEMA + tableDDLOptions;
+        INDEX_DDL =


Any risks associated with changing the static member? If there are no real benefits to a static member, perhaps we should make this a function local variable to avoid surprises.

It is actually used as part of a coprocessor test class in this file. But the tests that use that coprocessor is disabled. So I think it is ok to make them local

Ok .. We can perhaps do this work in another PR. Using statics makes it impossible to run these tests in parallel without worrying about updates happening concurrently.

priyankporwal · 2019-06-11T23:34:12Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

@@ -107,15 +121,17 @@ public static void doSetup() throws Exception {
        Map<String, String> clientProps = Maps.newHashMapWithExpectedSize(2);
        clientProps.put(QueryServices.TRANSACTIONS_ENABLED, "true");
        clientProps.put(QueryServices.INDEX_POPULATION_SLEEP_TIME, "15000");
+        clientProps.put(QueryServices.INDEX_REGION_OBSERVER_ENABLED_ATTRIB,


Nit: Consistent inlining of this put with others right above.

priyankporwal · 2019-06-11T23:34:58Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+            conn.setAutoCommit(true);
+            int num = 1;
+            createTableAndIndexForConsistentIndex(conn, fullTableName, fullIndexName, num);
+           // TestUtil.waitForIndexRebuild(conn, indexName, PIndexState.ACTIVE);


Please remove commented out code.

priyankporwal · 2019-06-12T16:50:52Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+        TABLE_NAME = fullTableName;
+        try (Connection conn = DriverManager.getConnection(getUrl(), props)) {
+            conn.setAutoCommit(true);
+            int num = 1;


nit: s/num/numRows

priyankporwal · 2019-06-12T16:51:51Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+        String indexName = "IND_" + generateUniqueName();
+        String fullTableName = SchemaUtil.getTableName(TestUtil.DEFAULT_SCHEMA_NAME, tableName);
+        String fullIndexName = SchemaUtil.getTableName(TestUtil.DEFAULT_SCHEMA_NAME, indexName);
+        TABLE_NAME = fullTableName;


Same concern here with changing static member field

priyankporwal · 2019-06-12T16:53:51Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+            assertTrue(rs.next());
+            assertEquals(num-1, rs.getInt(1));
+
+            // Disable data table so that delete fails on data table but index table row remains as unverified


Nice test! simulating the unverified index row

priyankporwal · 2019-06-12T16:55:43Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

@@ -24,14 +24,19 @@
 import static org.apache.phoenix.monitoring.GlobalClientMetrics.GLOBAL_MUTATION_COMMIT_TIME;
 import static org.apache.phoenix.query.QueryServices.WILDCARD_QUERY_DYNAMIC_COLS_ATTRIB;
 import static org.apache.phoenix.query.QueryServicesOptions.DEFAULT_WILDCARD_QUERY_DYNAMIC_COLS_ATTRIB;
+import static org.apache.phoenix.schema.types.PDataType.FALSE_BYTES;
+import static org.apache.phoenix.schema.types.PDataType.TRUE_BYTES;


I don't know much about this class yet. Bit with 492 lines of code change, I see no corresponding test changes. Are there no existing unit-tests that need changes? If not, should we start writing them with this change?

There is a MutationStateIT class, it might count as unit/it tests.
My changes are mostly refactors. There was a huge send() method, I refactored and moved some part of send code into sendMutations function. This is the place where client creates and sends mutations for index and data table.

priyankporwal · 2019-06-12T16:58:51Z

phoenix-core/src/main/java/org/apache/phoenix/util/IndexUtil.java

+        try {
+            desc = connection.getQueryServices().getTableDescriptor(index.getBytes());
+        } catch (SQLException ex) {
+            //logger.warn("Error during getTableDescriptor ", ex);


Can we throw this exception instead of silently ignoring it?

kadirozde · 2019-06-12T17:34:05Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

@@ -17,10 +17,15 @@
 */
 package org.apache.phoenix.end2end.index;

+import static org.apache.phoenix.end2end.IndexToolIT.assertExplainPlan;
+import static org.apache.phoenix.schema.types.PDataType.FALSE_BYTES;
+import static org.apache.phoenix.schema.types.PDataType.TRUE_BYTES;


Instead of using TRUE_BYTES and FALSE_BYTES, please use IndexRegionObserver.UNVERIFIED_BYTES and IndexRegionObserver.VERIFIED_BYTES. I have changed this recently.

kadirozde · 2019-06-17T17:04:38Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+            rs = conn.createStatement().executeQuery("SELECT COUNT(*) FROM " + fullIndexName);
+            assertTrue(rs.next());
+            assertEquals(numRows, rs.getInt(1));
+            verifyRowsForEmptyColValue(conn, fullIndexName, IndexRegionObserver.VERIFIED_BYTES);


The return value of verifyRowsForEmptyColValue() is not checked here and in the other places. Is it intentional? If so, what is the purpose of calling verifyRowsForEmptyColValue().

good catch!

Does anyone know if there is a compile time check for such return-value usages (unless there is explicit ignores)?

kadirozde · 2019-06-17T17:12:10Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+                            "EXPLAIN SELECT long_pk, varchar_pk, long_col1 FROM " + TABLE_NAME + " WHERE varchar_pk='varchar2' AND long_pk=2");
+            String actualExplainPlan = QueryUtil.getExplainPlan(rs);
+            assertExplainPlan(false, actualExplainPlan, fullTableName, fullIndexName);
+            assertFalse(rs.next());


The test has failed to delete the rows WHERE varchar_pk='varchar2'. What is the purpose of this select with WHERE varchar_pk='varchar2' AND long_pk=2? and why to expect assertFalse(rs.next())?

fixing it. The tests makes sure we use index for that select. rs.next should actually test the select and see it is not returning data.

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

kadirozde · 2019-06-18T04:26:44Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

                // build map from physical table to mutation list
                boolean isDataTable = true;
                while (mutationsIterator.hasNext()) {
                    Pair<PName, List<Mutation>> pair = mutationsIterator.next();
                    PName hTableName = pair.getFirst();
                    List<Mutation> mutationList = pair.getSecond();
+                    if (clonedDataMutations.size() == 0) {


clonedDataMutations are not used anymore. We can remove it.

kadirozde · 2019-06-18T04:33:08Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+            filterIndexCheckerMutations(physicalTableMutationMap, unverifiedIndexMutations,
+                    verifiedOrDeletedIndexMutations);
+
+            // Phase 1: Send verified indexes with VERIFIED=false


I suggest rephrasing the comment, for example:
// Phase 1: Send index mutations with the empty column value = "unverified"

kadirozde · 2019-06-18T04:35:44Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+                    physicalTableMutationMap.entrySet().iterator();
+            sendMutations(mutationsIterator, span, indexMetaDataPtr);
+
+            // Phase 3: Send verfied indexes with VERIFIED = true


I suggest rephrasing the comment, for example:
// Phase 3: Send put index mutations with the empty column value = "verified" and/or delete index mutations

priyankporwal · 2019-06-17T19:49:07Z

phoenix-core/src/test/java/org/apache/phoenix/util/TestUtil.java

@@ -957,7 +955,8 @@ public static void waitForIndexState(Connection conn, String fullIndexName, PInd
                }
            }
        } while (++nTries < maxTries);
-        fail("Ran out of time waiting for index state to become " + expectedIndexState);
+        fail("Ran out of time waiting for index state to become " + expectedIndexState + " last seen actual state is " +
+                (actualIndexState == null ? "" : actualIndexState.toString()));


Nit: Perhaps "Unknown" instead of ""?

priyankporwal · 2019-06-17T20:00:21Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+            rs = conn.createStatement().executeQuery("SELECT COUNT(*) FROM " + fullIndexName);
+            assertTrue(rs.next());
+            assertEquals(numRows, rs.getInt(1));
+            verifyRowsForEmptyColValue(conn, fullIndexName, IndexRegionObserver.VERIFIED_BYTES);


Does anyone know if there is a compile time check for such return-value usages (unless there is explicit ignores)?

priyankporwal · 2019-06-17T20:06:58Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+    private void createTableAndIndexForConsistentIndex(Connection conn, String tableName, String indexName, int numOfRowsToInsert)
+            throws Exception {
+        String ddl = "CREATE TABLE " + TABLE_NAME + TestUtil.TEST_TABLE_SCHEMA + tableDDLOptions;
+        INDEX_DDL =


Ok .. We can perhaps do this work in another PR. Using statics makes it impossible to run these tests in parallel without worrying about updates happening concurrently.

priyankporwal · 2019-06-18T16:33:02Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+                    verifiedOrDeletedIndexMutations);
+
+            // Phase 1: Send verified indexes with VERIFIED=false
+            // addRowMutations generates the mutations with VERIFIED=false


Is the comment with "addRowMutations" relevant here? If it is, I see that getting called from send() - so is it modifying the unverifiedIndexMutations iterator?

priyankporwal · 2019-06-18T16:34:00Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+            Iterator<Entry<TableInfo, List<Mutation>>>
+                    mutationsIterator =
+                    physicalTableMutationMap.entrySet().iterator();
+            sendMutations(mutationsIterator, span, indexMetaDataPtr);


Nit: For readability, we can simply make this sendMutations(physicalTableMutationMap.entrySet().iterator(), span, indexMetaDataPtr) similar to Phase1 & Phase3 send calls.

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

priyankporwal · 2019-06-18T16:35:49Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+                sendMutations(verifiedOrDeletedIndexMutations.entrySet().iterator(), span, indexMetaDataPtr);
+            } catch (SQLException ex) {
+                LOGGER.warn(
+                        "Ignoring exception that happened during setting index verified value to verified=TRUE",


Does this log-line need table/index names or row-key?

@gokceni - I agree with @priyankporwal , we should log the table and index names

priyankporwal · 2019-06-18T16:51:06Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+                    } else {
+                        // These mutations already have Unverified_Bytes set.
+                        addToMap(unverifiedIndexMutations, tableInfo, m);
+                        // We will need to have verified byte set version for the Phase 3.


Nit: "We will need .." sounds like that is being done somewhere else. Perhaps rephrase as "set verified bytes for phase3"

priyankporwal · 2019-06-18T16:52:15Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+
+                List<Mutation> mutations = pair.getValue();
+
+                if (mutations.size() == 0) {


Remove this size check? Won't the for loop be sufficient?

kadirozde

Nice work!

gjacoby126

@gokceni - As others have said, good job on the MutationState cleanup. I found mostly small stuff, but please see my comment about the perf of frequent checking for GlobalIndexChecker coprocessor.

gjacoby126 · 2019-06-24T20:39:30Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java

+
+    private void createAndPopulateTableAndIndexForConsistentIndex(Connection conn, String tableName, String indexName, int numOfRowsToInsert)
+            throws Exception {
+        String ddl = "CREATE TABLE " + TABLE_NAME + TestUtil.TEST_TABLE_SCHEMA + tableDDLOptions;


@gokceni - passing in a tableName parameter but not using it here, since the create statement is using the static value. Even if we need the static value to be set for some other reason, (and I agree with @priyankporwal that we ideally shouldn't need that), within this function we should be consistent.

gjacoby126 · 2019-06-24T20:58:47Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+                sendMutations(verifiedOrDeletedIndexMutations.entrySet().iterator(), span, indexMetaDataPtr);
+            } catch (SQLException ex) {
+                LOGGER.warn(
+                        "Ignoring exception that happened during setting index verified value to verified=TRUE",


@gokceni - I agree with @priyankporwal , we should log the table and index names

gjacoby126 · 2019-06-24T21:05:25Z

phoenix-core/src/main/java/org/apache/phoenix/query/ConnectionQueryServicesImpl.java

@@ -911,7 +911,8 @@ private void addCoprocessors(byte[] tableName, TableDescriptorBuilder builder, P
                    QueryServicesOptions.DEFAULT_INDEX_REGION_OBSERVER_ENABLED);

            if (tableType == PTableType.INDEX && !isTransactional) {
-                if (globalIndexerEnabled && !newDesc.hasCoprocessor(GlobalIndexChecker.class.getName())) {
+                if (globalIndexerEnabled && !newDesc.hasCoprocessor(GlobalIndexChecker.class.getName()) &&
+                        !isLocalIndexTable(builder.build().getColumnFamilyNames())) {


nit: can use newDesc for the TableDescriptor rather than using builder.build() to reconstruct it.

gjacoby126 · 2019-06-24T21:12:24Z

phoenix-core/src/main/java/org/apache/phoenix/util/IndexUtil.java

+
+    public static boolean isGlobalIndexCheckerEnabled(PhoenixConnection connection, PName index)
+            throws SQLException {
+        TableDescriptor desc = desc = connection.getQueryServices().getTableDescriptor(index.getBytes());


This is called from generateIndexData which will be high-traffic. What are the perf implications of grabbing the table descriptor very frequently, and do we need caching?

Looks like this creates an HTable each time which will in turn make a remote call to the Master each time
@gokceni @kadirozde

adding to a static map.

Actually I am not sure about caching this. If upgrade tool adds the coprocessor, map will go stale.

gjacoby126 · 2019-06-24T21:14:43Z

phoenix-core/src/test/java/org/apache/phoenix/util/TestUtil.java

+        return getRawRowCount(table, true);
+    }
+
+    public static int getRawRowCount(Table table, boolean isRaw) throws IOException {


nit: Seems a little odd to have a function named getRawRowCount that doesn't do a raw scan if you pass in false. (As opposed to having a getRowCount funciton with an isRaw parameter.)

gokceni · 2019-06-24T22:22:50Z

@gjacoby126 for your logging comment, it will complicate the code to get index name and table name at that part of the code. I can print the hashmap though

phoenix-core/src/main/java/org/apache/phoenix/util/IndexUtil.java

kadirozde · 2019-06-27T20:38:16Z

phoenix-core/src/main/java/org/apache/phoenix/schema/PTableImpl.java

@@ -1539,6 +1539,7 @@ public IndexType getIndexType() {
    public static PTable createFromProto(PTableProtos.PTable table) {
        if (table==null)
            return null;
+


PTableImpl.java should be removed from this patch as it is included by mistake

kadirozde

+1

gjacoby126

@gokceni - Some nits + a bug when the filter logic encounters an Append mutation, which should be a simple fix.

gjacoby126 · 2019-06-27T20:57:12Z

phoenix-core/src/main/java/org/apache/phoenix/util/IndexUtil.java

@@ -137,7 +143,12 @@
 public class IndexUtil {
    public static final String INDEX_COLUMN_NAME_SEP = ":";
    public static final byte[] INDEX_COLUMN_NAME_SEP_BYTES = Bytes.toBytes(INDEX_COLUMN_NAME_SEP);
-
+
+    private static Cache<String, Boolean> IndexNameGlobalIndexCheckerEnabledMap = CacheBuilder.newBuilder()


super tiny nit: variables in Java should start with a lower case letter. Only bother with this if you need to make a change for other reasons. :-)

gjacoby126 · 2019-06-27T20:59:33Z

phoenix-core/src/main/java/org/apache/phoenix/util/IndexUtil.java

+                }
+            }
+            IndexNameGlobalIndexCheckerEnabledMap.put(indexName, result);
+        } catch (TableNotFoundException ex) {


If you can pass in the PTable rather than the PName, you could interrogate the PTable to see whether it's a view index or not and save yourself the RPC and the exception

The thing is I don't have PTable, I have just the name. It is in that TableRef object. There is a ptable but it is the base table

gjacoby126 · 2019-06-27T21:36:59Z

phoenix-core/src/main/java/org/apache/phoenix/execute/MutationState.java

+                        // The Delete gets marked as unverified in Phase 1 and gets deleted on Phase 3.
+                        addToMap(unverifiedIndexMutations, tableInfo, put);
+                        addToMap(verifiedOrDeletedIndexMutations, tableInfo, m);
+                    } else {


@gokceni - This will fail with a ClassCastException if it encounters an Append, a third kind of Mutation that acts somewhat like a Put, but reads the existing value and increments / concatenates to it.

Since Appends also have an addColumn method, you probably just need another else if for them and similar logic to the Put.

Oh, and Increment.

According to @kadirozde , "Increment and append should not be supported by Immutable tables by definition
We need to support them for mutable tables". I guess he will have a jira for that.

priyankporwal · 2019-06-27T22:48:08Z

phoenix-core/src/it/java/org/apache/phoenix/end2end/IndexToolForPartialBuildIT.java

+                    hTable.delete(new Delete(res.getRow()));
+                }
+            }
+            hTable.close();


Please create hTable with try (i.e. try with resources)

…l Tables

gjacoby126

+1

swaroopak reviewed Jun 12, 2019

View reviewed changes

priyankporwal suggested changes Jun 12, 2019

View reviewed changes

kadirozde reviewed Jun 12, 2019

View reviewed changes

gokceni force-pushed the PHOENIX-5211 branch from 30dbcdf to 93bafe9 Compare June 14, 2019 00:06

kadirozde reviewed Jun 17, 2019

View reviewed changes

gokceni force-pushed the PHOENIX-5211 branch from 93bafe9 to f8aaff2 Compare June 17, 2019 23:09

kadirozde reviewed Jun 18, 2019

View reviewed changes

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java Show resolved Hide resolved

kadirozde reviewed Jun 18, 2019

View reviewed changes

phoenix-core/src/it/java/org/apache/phoenix/end2end/index/ImmutableIndexIT.java Show resolved Hide resolved

kadirozde reviewed Jun 18, 2019

View reviewed changes

priyankporwal suggested changes Jun 18, 2019

View reviewed changes

gokceni force-pushed the PHOENIX-5211 branch 3 times, most recently from 1c6679f to 5b3c9fc Compare June 18, 2019 18:33

kadirozde approved these changes Jun 19, 2019

View reviewed changes

gjacoby126 requested changes Jun 24, 2019

View reviewed changes

gokceni force-pushed the PHOENIX-5211 branch from 5b3c9fc to 0af2863 Compare June 25, 2019 18:25

gokceni commented Jun 25, 2019

View reviewed changes

phoenix-core/src/main/java/org/apache/phoenix/util/IndexUtil.java Outdated Show resolved Hide resolved

gokceni force-pushed the PHOENIX-5211 branch 9 times, most recently from 218e2fd to 29fd078 Compare June 26, 2019 17:48

gokceni force-pushed the PHOENIX-5211 branch 2 times, most recently from b52a78e to 2c50607 Compare June 27, 2019 20:27

kadirozde reviewed Jun 27, 2019

View reviewed changes

gokceni force-pushed the PHOENIX-5211 branch from 2c50607 to 2fb57ee Compare June 27, 2019 20:41

kadirozde approved these changes Jun 27, 2019

View reviewed changes

gjacoby126 requested changes Jun 27, 2019

View reviewed changes

priyankporwal reviewed Jun 27, 2019

View reviewed changes

PHOENIX-5211 Consistent Immutable Global Indexes for Non-Transactiona…

bca34c8

…l Tables

gokceni force-pushed the PHOENIX-5211 branch from 2fb57ee to bca34c8 Compare June 27, 2019 23:55

gjacoby126 approved these changes Jun 28, 2019

View reviewed changes

kadirozde closed this Jul 3, 2019


		List<Mutation> mutations = pair.getValue();

		if (mutations.size() == 0) {

PHOENIX-5211 Consistent Immutable Global Indexes for Non-Transactiona… #517

PHOENIX-5211 Consistent Immutable Global Indexes for Non-Transactiona… #517

Conversation

gokceni commented Jun 11, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kadirozde left a comment

Choose a reason for hiding this comment

gjacoby126 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gokceni commented Jun 24, 2019

Choose a reason for hiding this comment

kadirozde left a comment

Choose a reason for hiding this comment

gjacoby126 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gjacoby126 left a comment

Choose a reason for hiding this comment