Munin print-screen for reference: http://img29.imageshack.us/i/munin.gif/
I monitor a web-server which basically runs apache+php+mysql. No emails, no FTP, pure web-server.
The server is serving pages slow...
I think the bottleneck is MySQL configuration.
- most of queries are SELECTs, very few INSERTs and rare DELETEs
- mirror backup runs around 3AM - represented by the red line at the image provided
- I use tuning-primer.sh to help me adjust MySQL
- Server uptime is 26 days
- 2GB ram, Pentium(R) Dual-Core E5200 @ 2.50GHz, 7200 disk
- I have log_slow_queries enabled and long_query_time = 1
- Today I switched on the log-queries-not-using-indexes. That's why the slow-queries graph is displaying a peek
MyQSL tweaks I did on the my.cnf:
- join_buffer_size = 10M
- max_heap_table_size = 200M
- tmp_table_size = 300M
- init-connect = 'SET NAMES utf8; SET CHARACTER SET utf8;'
- default-character-set = utf8
- character-set-server = utf8
- open_files_limit = 2048
- key_buffer = 256M
- max_allowed_packet = 16M
- thread_stack = 192K
- thread_cache_size = 8
max_connections = 100
- table_cache = 1024
thread_concurrency = 10
- query_cache_limit = 4M
- query_cache_size = 64M
Why yesterday and before everything was working OK? We didn't add new code or having a peek of visitors.
By looking at munin's graph do you see anything odd beside the slow-queries peek?
Do you recommend installing other tools in order to improve monitoring?