MariaDB
Contents
PHP/MySQL med failover
- Alle der udvikler PHP til info SKAL læse
Monitorering
Galera SYNC
Følgende targets skal sættes i nagios for alle 3 nodes
- Galera cync
Intel SSDSW RaidMySQL connection- CPU forbrug
- free space:
root partition/mnt/data
Ting at være obs på
- LOAD DATA INFILE - forventer lokale filer
- LOAD DATA LOCAL INFILE er implicit IGNORE, så der kommer ingen fejl hvis at der er fejl i datafilen
- auto_increment tabeller laver ikke længere nummerserien som direkte fortløbende (1,2,3) men springer derimod i numrene (3,7,10)
- genereringen af SSCC stregkoder samt sms koder forudsætter direkte sekvens - disse skal kodes om.
- læs mere...
- Galera Limitations Alle skal Læse denne (vær især obs på "Transaction Size" sektionen)
Galera
- https://mariadb.org/installing-mariadb-galera-cluster-on-debian-ubuntu/
- https://mariadb.com/kb/en/mariadb/getting-started-with-mariadb-galera-cluster/
husk at sætte
wsrep_node_name="node-name" wsrep_on=ON
den sidste er yderst vigtigt for at galera vil starte
Links
- Percona Toolkit
- Generic:
- Migration:
Checksum
Note: Nuværende slave data er bygget op fra mysqldump - som skipper dato stemplede tabeller, så når de er fraværende vil pt-table-checksum brokke sig
Gennemført på
- test
- master
- logs
06-12T21:36:31 Skipping table logs.morgen_rapport because on the master it would be checksummed in one chunk but on these replicas it has too many rows: 139820 rows on info-backup 142723 rows on MariaDB-01 The current chunk size limit is 137222 rows (chunk size=68611 * chunk size limit=2.0).
- fulddaeking
- produktion_status
06-12T22:08:55 Skipping table produktion_status.30-5lossalgnyture_tan_frem30052013 because it has problems on these replicas: Table produktion_status.30-5lossalgnyture_tan_frem30052013 does not exist on replica info-backup Table produktion_status.30-5lossalgnyture_tan_frem30052013 does not exist on replica MariaDB-01 06-12T22:13:19 0 0 10741793 144 0 260.846 produktion_status.aflo 06-12T22:13:25 Skipping chunk 1 of produktion_status.afregning because MySQL used only 7 bytes of the PRIMARY index instead of 134. See the --[no]check-plan documentation for more information.
Centos/RHEL
https://github.com/Tunts/bakery/wiki/Setting-up-MariaDB-10.1---Galera-Cluster-on-Centos-7
https://www.percona.com/doc/percona-xtrabackup/LATEST/installation/yum_repo.html
"Due to a bug in socat on CentOS 7, the xtrabackup-v2 method does not work. This can be solved by installing an older version of socat or by using the rsync method for wsrep_sst_method"
remember
- (define your cluster as a set of hostnames and NOT as a list of IP's) ?
- If using xtrabackup transfer
- socat in centos 7/rehl 7 doesn't work with SST - so install anothter one
- to set
wsrep_sst_method=xtrabackup-v2
- and set wsrep_sst_auth="user:secret" [1]
- and create the user
- mysql@percona1> CREATE USER 'sstuser'@'localhost' IDENTIFIED BY 's3cret';
- mysql@percona1> GRANT PROCESS, RELOAD, LOCK TABLES, REPLICATION CLIENT ON *.* TO 'sstuser'@'localhost';
- mysql@percona1> FLUSH PRIVILEGES;
- disable selinux !!!!
- disable host firewall totally
- systemctl stop firewalld
- systemctl disable firewalld