我正在尝试在 Spring Boot 服务中将一个大数据库从 derby 迁移到 HSQLDB，例如 10 列的几个表中的 1.5M regs。我正在检查 VisualVM； byte 和 char 消耗大量内存。但时间上最大的增量是在 Derby 赛中。
有时错误是在此处抛出，但有时是在其他 Controller 中抛出。我不想触摸所有文件来添加我的 catchOutofMemory 以重新启动。
以下是我的代码版本， block 注释显示了流程的恢复:
run(){//thread inside static function.
while(keepMigrating){
keepMigrating=Migrate();
}
}
private static boolean Migrate(JdbcTemplate derby,JdbcTemplate hsql){
int regs = 100000;
PreparedStatement statement = null;
ResultSet rs = null;
PreparedStatement statementHSQL = null;
try {
for (String table : tables) {//tables contains all tables to migrate
//check how many registers left asd asign to cant, if cant is 0 the empty is true.
PreparedStatement statementUpd[];
while (!empty) {
if (trys <= 0) throw new Exception("redo");
//check how many registers left asd asign to cant, if cant is 0 the empty is true and out of bucle and ready to check next table
/*
*Next process resume as:
*fetch data from derby that hasnt been migrated limited by cant
*create a batch to insert in hsql
*create a update for derby
*create a delete in case someting goes wrong
*excecute insert and update, if someting in batch fail delete the entry in migrate table
*reduce regs to get out of migrate method at some ponint.
*/
statement = derby.getDataSource().getConnection().prepareStatement(
MessageFormat.format(select_all_migrate_false_and_fetch_cant,table));
statementUpd = new PreparedStatement[cant];
ArrayList<String> deleteIds = new ArrayList<>();
StringBuilder columnNames = new StringBuilder();
StringBuilder updateSQL = new StringBuilder();
StringBuilder bindVariables = new StringBuilder();
try {
ResultSetMetaData meta = rs.getMetaData();
for (int i = 1; i <= meta.getColumnCount(); i++) {
if (!meta.getColumnName(i).equals("MIGRATED")) {
if (i > 1) {
columnNames.append(", ");
bindVariables.append(", ");
}
columnNames.append(meta.getColumnName(i));
bindVariables.append('?');
}
}
String sql = "INSERT INTO " + table.substring(4) + " ("
+ columnNames
+ ") VALUES ("
+ bindVariables
+ ")";
statementHSQL = hsql.getDataSource().getConnection().prepareStatement(sql);
HashMap<String, Object> data = new HashMap<>();
int row = 0;
int lastId = 0;
String columnName;
while (rs.next()) {
for (int i = 1; i <= meta.getColumnCount(); i++) {
columnName = meta.getColumnName(i);
Object o = rs.getObject(i);
statementHSQL.setObject(i, o);
if (columnName.equals(mainColumn))
deleteIds.add(String.valueOf(o));
if (!(meta.getColumnType(i) == 2004)) data.put(columnName, o);
if (columnName.equals(mainColumn)) id = rs.getObject(i);
}
int c = 1;
String update = MessageFormat.format("INSERT INTO {0}M ({1}M, MIGRATED) VALUES(?, TRUE)",
table.substring(4), mainColumn).replace("\"M", "M\"");//migrated state is saved in other table
lastId = Integer.valueOf(String.valueOf(id));
statementUpd[row] = derby.getDataSource().getConnection().prepareStatement(update);
statementUpd[row].setObject(1, rs.getObject(mainColumn));
updateSQL = new StringBuilder();
statementHSQL.addBatch();
row += 1;
}
/*
* Build delete query in case of inserted values in HSQLDB but not updated in DERBY
*/
StringBuilder builder = new StringBuilder();
builder.append("(");
int count = 1;
for (String s : deleteIds) {
if (count > 1) builder.append(", ");
builder.append("?");
count++;
}
builder.append(")");
String str = builder.toString();
String queryDelete = "DELETE FROM " + table.substring(4) + " WHERE " + mainColumn + " IN " + str;
PreparedStatement statementHSQLDel = hsql.getDataSource().getConnection().prepareStatement
(queryDelete);
int c = 1;
for (String s : deleteIds) {
statementHSQLDel.setObject(c, s);
c++;
}
boolean deletes = statementHSQLDel.execute();
statementHSQLDel.close();
try {
DatabaseUtils.close(statementHSQLDel);
} catch (Exception e) {
catchOutOfMemory(e);
}
int[] result = statementHSQL.executeBatch();
StringBuilder resultS = new StringBuilder();
int stCounter = 0;
int stCounterInsert = 0;
int stCounterUpdate = 0;
String notarydebug;
for (int i : result) {
int upd = 0;
try {
if (i == 1) upd = statementUpd[stCounter].executeUpdate();
} catch (Exception e) {
catchOutOfMemory(e);
}
stCounterInsert += i;
stCounterUpdate += upd;
resultS.append(",").append(String.valueOf(i)).append("-").append(String.valueOf(upd));
stCounter += 1;
}
statementHSQL.clearBatch();
try {
DatabaseUtils.close(statementHSQL);
} catch (Exception e) {
catchOutOfMemory(e);
}
} catch (SQLException se) {
catchOutOfMemory(se);//otherstuff
} catch (Exception e) {
catchOutOfMemory(e);
}
try {
DatabaseUtils.close(rs);
DatabaseUtils.close(statement);
} catch (Exception e) {
catchOutOfMemory(e);
}
regs=regs-cant;
}
}
}catch (Exception e) {
if (e.getMessage().equals("redo")) return true;//end the loop of regs maximun and get out of method.
}
return false;//end migration succesfully
}
private static int catchOutOfMemory(Throwable e) {
if (e == null) return 0;
if (e instanceof OutOfMemoryError) {
Application.restartBat();
return 1;
} else {
return catchOutOfMemory(e.getCause());
}
}
编辑: 所以我按照评论中的建议进行了更改以接受提交，如下所示:
Connection hsqlCon;
PrepareStatement hsqlStm;
hsqlCon = JdbcHSQLDB.getDataSource().getConnection();
hsqlStm = hsqlCon.prepareStatement(sql);
hsqlStm.addBatch();
hsqlStm.execute();
hsqlStm.close();
hsqlCon.close();
但我得到了相同的堆内存消耗:
最佳答案
HSQLDB 中的表类型从提供的代码中看不清楚。您必须对每个表使用此语句一次，以确保表数据存储在 finename.data 文件中:
SET TABLE tableName TYPE CACHED
报告的批处理 INSERT 顺序不正确。使用此序列:
Connection hsqlCon;
PrepareStatement hsqlStm;
hsqlCon = JdbcHSQLDB.getDataSource().getConnection();
hsqlStm = hsqlCon.prepareStatement(sql);
{ // repeat this block until all is finished
{ // repeat for 1000 rows
hsqlStm.addBatch();
}
hsqlStm.executeBatch(); // after every 1000 rows
}
hsqlStm.close();
hsqlCon.close();
关于java - 将 Big Derby 嵌入式数据库迁移到 HSQLDB 抛出 java.lang.OutOfMemoryError : Java heap space，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/53200371/