Solution of gbase 8s livelock and deadlock

         The lock problem ( Lock wait 、 Deadlock ) Most of them do not understand the management of locks correctly 、 The way of use leads to . To solve the problem of lock , We can try from the following aspects .

        1.GBase8s  database Lock resource Set up ONCONFIG Parameters LOCKS Setting is to configure an appropriate number of locks for the database . By searching online Does the log dynamically add locks ：grep 'dynamically allocated' online.log Set a reasonable locking mode for the table 、 Row level lock or page level lock .

        2. Application lock Usage mode Set up adopt onstat -g sql see SESSION Lock waiting mode and isolation level used , Set a reasonable isolation level and lock waiting mode for different applications . If the following statement is added to the application ：

        set isolation to dirty read;

        set lock

        3． Applications avoid using unnecessary locks Many applications SQL Statement because the index is not used correctly INDEX, This results in locking the entire table , The problem of lock concurrency is caused by . Need to pass through SQL Optimize and create reasonable indexes , Avoid lock conflicts caused by sequential scanning .

         Lock waiting problem due to sequential scanning ： Create table test_lock, Using row level locks , And in c1 Create index on field , Insert 3 Line test data .

        create table test_lock (c1 int,c2 int,c3 char(10)) lock mode row;

        create index idx_test_lock on test_lock(c1);

        insert into test_lock values(1,1,'abc');

        insert into test_lock values(2,2,'abc');

        insert into test_lock values(3,3,'abc');

        Session 1 Modify the... By index scanning 2 rows ：   

      Begin work;

        update test_lock set c3='new' where c1=2;

         here ,Session 1 Add... To the second line of record X lock .

        Session 2 adopt c2 Fields are scanned sequentially to modify the 3 rows . 

       Begin work;

        update test_lock set c3='new' where c2=3;

         because Session 2 Adopt sequential scanning mode , In the course of scanning, it is found that 2 Row record added X lock , Therefore, a lock conflict error is prompted ：

        244: Could not do a physical-order read to fetch next row.

        107: ISAM error: record is locked.

         If in c2 Add index... To the field , Or by c1=3 Modify the record in the third line , Then there will be no lock problem .

        4． Applications Optimize to avoid too long Occupancy lock A lock is a public resource , Lock resources need to be released as soon as possible , Reduce waiting , If every transaction can be completed at the microsecond level , Then the lock will not be our concern , Reducing the lock time is also the key to solve the lock problem . We can go through Avoid big business , Split a large transaction into multiple submissions , So the lock can be released as soon as possible . in addition , Optimize application processing performance , Such as the rational use of indexes , Can speed up transaction processing , The possibility that lock conflicts can be solved invisibly .

        5． Set up Reasonable isolation level Different isolation levels use different locks , Set a reasonable isolation level , It can improve the concurrency of applications , Improve the processing capacity of the system , Reduce lock conflicts , Avoid lock problems caused by read operations .