1600NVPS - scaling to 3000NVPS

[thesubmitter]

Hi,

I am in the process of replacing a zabbix environment with 1600NVPS (v3. 4)and retaining only 10 days of data to a system that should store at least 90 days of history, 1 year of events nad 3 years of trends (v5). I also want to be able to handle 3000NVPS if necessary.

Money isn't really an issue but I want to be close to the mark.

I'm planning for:
6TB SSD (4x3.2TB) with HP SmartArray MR BBC Card either Raid 10 or Raid 5
1 Xeon Gold 6230R
1 256GB Ram (I have money)


-I am planning to use timeseriesDB
-I am not planning for HA (or if I have a backup it will be asynchronous and a DR failover type)
-I will have one machine as DB and Server. I think at the specs I am putting up I should be fine and it eliminates network latency issues.


Am I missing somehting....

Thanks in Advance!

[vendrusculo]

Hi!

Looks like you are buying a bazooka to kill ants

I have 7900+ vps, 3 days of history (minute), 1 year of trends (hourly) and using a postgres with partitioned tables (less efficient than timescaleDB) and the problem is always on the database! The database size today is 1.1TB

Zabbix by it self will not be a problem with small machines even using virtual machines or docker, probably you will need at maximum a proxy to keep your queue clean...

I would say to be prepare to handle the database needs instead spend too much time on Zabbix


Thanks,
Leo

[thesubmitter]

The server will be my DB too and I will add as many proxies as needed