Huge DB Server requirements?

[astyanax]

Hello,

I have noticed that by default, statistics in Zabbix are set to be kept for only 7 days. We are looking to do more resource utilization and require the data to be kept for 2 years. I have been doing some analysis, and have noticed that with only a portion (36) of our hosts we wish to monitor, with an update interval of 5 minutes, and checking 27 items per host, the databse still grows by almost 7.5MB per day! I was looking into this further, and can only assume it is because the history records are each a double(16,4), which I imagine take up quite a bit of disk space per record. My question is will Zabbix function will a different history.value record size? Can the precicion be reduced without screwing up Zabbix? Besides a loss of precision, what else might be affected by changing this value?

[Alexei]

...with an update interval of 5 minutes, and checking 27 items per host, the databse still grows by almost 7.5MB per day!

7.5MB per day means 5.4GB in two years. I don't think this is huge database. ZABBIX is already used in environments with more than 50GB of data without significant performance hit.

Yes, you can decrease precision of history.value if you want. I do not think it will affect anything apart of the data precision

[LEM]

7.5MB per day means 5.4GB in two years. I don't think this is huge database.

We ran a zabbix-1.0 prototype environment (40 targeted servers, 3 month retention, checks every 5-30s for most of checks). Database size is about 3 Gb (using MySQL with innodb support configured). No performance issues, albeith the prototype run on a old hardware (intel P3/700Mhz, 512Mb RAM, EIDE disks).

We plan to use 1.0 (upgrade to 1.1 when will be stable) for about 200 targets (servers & network equipements), retention 395 days, checks every 60 s (average), about 25 checks per target (average). We plan a database size of 13 Gb at the end of the year (splitted on multiple innodb datafiles).

ZABBIX is already used in environments with more than 50GB of data without significant performance hit.

Interresting! Some more user testimonials should encourage some of us to go ahead using/extending zabbix@work.

[Alexei]

Interresting! Some more user testimonials should encourage some of us to go ahead using/extending zabbix@work.

Check this thread: http://www.zabbix.com/forum/showthre...?p=598#post598

Riegersteve's ZABBIX takes care of about 1000 servers, 300 cisco plus 200 printers. It turns into 20-22K of metrics being monitored. I was impressed!

[LEM]

Check this thread: http://www.zabbix.com/forum/showthre...?p=598#post598

Ouch! effectively! impressive ! We need an 'Hall Of Fame for this kind of zabbix plateform!

Riegersteve : what is the size of your zabbix database? how many (average) items/second do you trap/suck? what average retention time do you use?