Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
10.2.8
-
Centos 7
Description
After upgrade to MariaDB from 10.1 to 10.2.8.
We are using wsrep with 3 nodes cluster.
One node for Write and two nodes for read data.
sometimes I see several connections that stuck in Unlocking tables in Write node.
Even I kill them, they will be alive.
I don't know when it happen, but I know it maybe happen on different databases and not related to specific database or table.
my.cnf setting(this is same in all nodes)
#
|
# This group is read both both by the client and the server
|
# use it for options that affect everything |
#
|
[client-server]
|
|
#
|
# include all files from the config directory
|
#
|
!includedir /etc/my.cnf.d
|
|
[mysqld]
|
skip-name-resolve
|
datadir = /home/mysql/
|
tmpdir = /home/mysqltmp/
|
port = 3306 |
socket = /home/mysql/mysql.sock
|
|
# General #
|
log_error = /home/mysql/node1.err
|
default-storage-engine = innodb |
query_cache_limit = 0 |
query_cache_size = 0 |
query_cache_type = 0 |
bind-address = 0.0.0.0 |
max-connections = 500 |
max-allowed-packet = 15M
|
table_open_cache = 60000 |
tmp_table_size = 512M
|
|
#FTS
|
innodb_ft_result_cache_limit=256M
|
|
|
# Binlog #
|
binlog_format = ROW
|
sync_binlog = 0 |
|
# Galera #
|
wsrep_on=on
|
wsrep_provider=/usr/lib64/galera/libgalera_smm.so
|
wsrep_cluster_address="gcomm://xx.xx.xx.xx,xx.xx.xx.xx,xx.xx.xx.xx" |
wsrep_cluster_name='galera_cluster' |
wsrep_node_address='xx.xx.xx.xx' |
wsrep_node_name='node1' |
wsrep_provider_options="gcache.size=25G;gcs.fc_limit=4000;gcache.page_size=512M;gcs.fc_master_slave=YES;gcs.fc_factor=1.0;" |
wsrep_sst_method=xtrabackup-v2
|
wsrep_sst_auth=xtrabackup:1111111 |
|
|
#Replication
|
gtid_domain_id=1 |
server_id=1 |
log_slave_updates=1 |
log_bin=binlog
|
expire_logs_days=2 |
|
# Innodb #
|
innodb_file_format = BARRACUDA
|
innodb-flush-method = O_DIRECT
|
innodb_autoinc_lock_mode = 2 |
innodb_log_file_size = 1G
|
innodb_log_buffer_size = 512M
|
innodb_file_per_table = 1 |
innodb_flush_log_at_trx_commit = 0 |
innodb_buffer_pool_size = 100G
|
innodb_buffer_pool_instances = 20 |
innodb_write_io_threads = 16 |
innodb_read_io_threads = 16 |
innodb_change_buffering = all
|
transaction-isolation = READ-COMMITTED
|
|
[sst]
|
compressor="pigz" |
decompressor="pigz -d" |
Attachments
Issue Links
- is caused by
-
MDEV-13534 InnoDB STATS_PERSISTENT fails to ignore garbage delete-mark flag on node pointer pages
- Closed