I know I'm not the first one asking, but can't find the answer.
I have a postgresql db for my ( inherit ) zabbix server, that I'm trying to set up a bit better than my previous collegue.
At the moment the DB is huge compared to the number of items ( almost 500 hosts, 65000 items and 35000 trigger ) and before my intervention they use to keep month and month of history as it was not clear for them the differences between history and trends.
Now I've reduced the history retention ( in general ) to 7d and housekeeping process is doing it's work but....
... history table at the moment has 1.800.000.000+ rows ( that's almost two billion rows ) and in the select I'm doing, if I keep last month I should loose at least half of these.
At the moment the housekeeping delete ~8 milion rows per cycle, but it take quite a while so is not very efficient ( in the last two days I only "lost" 22million rows ).
Now the question: is it unsafe to manually delete the biggest part of these records to help housekeeping do his job ?
I don't know when zabbix compute their trends, and I don't want to loose trends data ( while I want to reduce history size ).
Thanks in advance.
Pierluigi
I have a postgresql db for my ( inherit ) zabbix server, that I'm trying to set up a bit better than my previous collegue.
At the moment the DB is huge compared to the number of items ( almost 500 hosts, 65000 items and 35000 trigger ) and before my intervention they use to keep month and month of history as it was not clear for them the differences between history and trends.
Now I've reduced the history retention ( in general ) to 7d and housekeeping process is doing it's work but....
... history table at the moment has 1.800.000.000+ rows ( that's almost two billion rows ) and in the select I'm doing, if I keep last month I should loose at least half of these.
At the moment the housekeeping delete ~8 milion rows per cycle, but it take quite a while so is not very efficient ( in the last two days I only "lost" 22million rows ).
Now the question: is it unsafe to manually delete the biggest part of these records to help housekeeping do his job ?
I don't know when zabbix compute their trends, and I don't want to loose trends data ( while I want to reduce history size ).
Thanks in advance.
Pierluigi
Comment