Database partioning with postgresql and partition manager or using timescaledb?

[Jason]

In moving up to version 5 of zabbix I'm in process of planning a migration from an old version 9 of postgres (using trigger based date partitioning on history and trend tables) on an old server to the current stable version 13 on a new database server. I've been trying to work out performance wise if it's better to use the native partitioning or use timescaledb. I've seen some reports that on 12 and 13 there is very little performance difference in it and if we're not keeping a massive amount of history then the advantages of compression are minimal.

Is there anything else I've missed that would mean that timescaledb offers other benefits that I won't get without it?

[[email protected]]

Hi Splitek,

I did not want to hijack this thread this this question however it may be useful for myself and others on the timescaledb topic....

I am using postgresql partitioning and backing up the public schema nightly and only dumping the individual trends partitions once after the day has finished (i.e. I did not back the complete database up every night as our DB is a TB).

How have you approached your backups with timescaledb?

Cheers
James

[[email protected]]

Hi Splitek,

Thanks for the response, I have just built a HA ProgreSQL setup with Repmgr, PgBouncer and HAProxy.

It did not occur to me to perform the backup on one of the replicas which will help out with load on the master.

Unfortunately for us we have SLA's to which will need data to substantiate if needed so I will have to look at some method and you have given me some tips - Thanks.

Cheers
James

[[email protected]]

Hi Cyber,

'Large' is very subjective...

For us our largest install is currently monitoring 10952 hosts, 3478637 items, 774856 triggers pushing 6500+ nvps (which I would consider medium)

Large I would be thinking in terms of just nvps, probably 10K+

Cheers
James

[cyber]

Just asking... What you consider "large scale"?...

[cyber]

I was just thinking - at which moment Timescale benefits can be seen or felt, at which moment it is large enough.....

[cyber]

We have it in use... no complaints... I just have some "existential dilemmas".. is everything ok, because we have it or did we add extra complexity... maybe it would have worked good enough without it also..
housekeeper pops up to 4-5% usage every hour for a minute or so .. in nvps scale, not something very big... 2.6k, close to 10k hosts, 500+k items 360k triggers..