Migrating to timescaleDB, what do I measure for data retention performance?

[jswheeler]

I've been running Zabbix on my home network for a few years now. I've become increasingly annoyed by the housekeeping process triggering an alarm. This is my one year trend of the housekeep utilization (below). I have a small number of devices that I monitor with both agents and SNMP.


Though the trend data is a bit misleading (the flaw of averages, my current trends below)

My ask, what can I measure, once I follow the conversion process to tell if things are better? The housekeeping process will no longer be enabled/used after the process is complete, but what metric or measure can I track to demonstrate improvement? This article was great and show metric ingestion (https://blog.zabbix.com/zabbix-time-...escaledb/6642/) but is there an operational metric that I can show to demonstrate the improvement or the health of the data retention system?

[cyber]

Housekeeper is still in use if you go to Timescale.. with TS you will override history settings and will be dropping data by chunks not by records (that was all written in that article also). So the housekeeper running time should be greatly reduced...

[jswheeler]

Do you think it's fair to say that the work of managing history shifts from Zabbix processes to DB processes (postgres in my case) so it's reasonable to see a drop in utilization on the zabbix server and a possible increase on the DB server?

[cyber]

Dropping a chunk from timescale table is less work than doing search and delete statements ... so there should be drop of utilization in both sides...
If your housekeeper run usualy takes 10m for example... it is doing all the searches and deleting etc.. going with TS, it does not need to do that, it just instructs DB to remove certain chunk(s). housekeeper should run much faster.