I started to have very high disk I/O which brought the server to be very slow (Ubuntu 18.04 LTS - 8GB -2xCPU - 80GB SSD - VM) with 256 devices reporting every 10 secs. DB is about 12GB with 54.000.000 records in tc_positions.
Suddenly most of the devices went to offline state.
Temporary solution was restart traccar . But few hours later , same problem again.
Installing atop showed that disk I/O was over 100% most of the time and started digging in mysql documentation.
https://dev.mysql.com/doc/refman/8.0/en/innodb-buffer-pool-resize.html
Finally added this to the /etc/mysql/mysql.conf.d/mysqld.cnf these 4 lines
innodb_flush_method = 1
innodb_buffer_pool_size =4G
innodb_buffer_pool_instances =8
innodb_buffer_pool_chunk_size =535822336
changing buffer pool size to 2G and buffer pool instances to 4 does not differ of the current configuration. maybe some more testing is required, but so far offline problem was gone.
the tool I use to check the device status is the following:
SELECT COUNT(*) FROM `tc_devices` WHERE lastupdate > NOW() - INTERVAL 5 MINUTE
So we this data we know if most of them are reporting and updating fine
I started to have very high disk I/O which brought the server to be very slow (Ubuntu 18.04 LTS - 8GB -2xCPU - 80GB SSD - VM) with 256 devices reporting every 10 secs. DB is about 12GB with 54.000.000 records in tc_positions.
Suddenly most of the devices went to offline state.
Temporary solution was restart traccar . But few hours later , same problem again.
Installing atop showed that disk I/O was over 100% most of the time and started digging in mysql documentation.
https://dev.mysql.com/doc/refman/8.0/en/innodb-buffer-pool-resize.html
Finally added this to the /etc/mysql/mysql.conf.d/mysqld.cnf these 4 lines
changing buffer pool size to 2G and buffer pool instances to 4 does not differ of the current configuration. maybe some more testing is required, but so far offline problem was gone.
the tool I use to check the device status is the following:
SELECT COUNT(*) FROM `tc_devices` WHERE lastupdate > NOW() - INTERVAL 5 MINUTESo we this data we know if most of them are reporting and updating fine