I did one final tune-up for my 1GB RAM Akkoma instance earlier today — I optimized the Posgres DB as per this Akkoma guide: https://docs.akkoma.dev/stable/configuration/postgresql/ and the help of PGTune: https://pgtune.leopard.in.ua/

Just adding a swap file (earlier) seemed to make things so much better and I can't say for certain that the Postgres tuning made a huge difference ... or not. But CPU load has certainly gone down a lot and the memory usage, while still substantial, is not always at a 100%. In fact, since the above changes, I don't believe memory usage has topped 90%.

So that's where it is after 24 hours of running the Akkoma server. This server incidentally has nothing else on it except for Akkoma 🙂

I'm adding some charts for the 24-hour period so that you can see how it's gone so far — where the lines plunge suddenly today is from around the time the swap file was added and the Postgres tuning done, one after the other with maybe 15 minutes interval in between.

#Akkoma #Ubuntu #Server #VPS
Graph showing CPU usage percent…
Graph showing memory usage perc…
Graph showing disk I/O for the …
Graph showing server load for t…

@fahim

Looks like your VPS has stopped thrashing due to OOM situations which is amazing considering your VPS only has 1 GB of RAM.

How much is pgsql taking up in terms of RAM? On my instance it's currently using 700MB...

@deltatux Is there a way to determine how much RAM Postgres is taking up? According to "top" and "ps" I have multiple Postgres processes and they show the % of memory used and the total of the percentage is higher than 100% 😛

@fahim

I installed my pgsql in a Docker container so it’s easier to manage. I just use this command to show how much resources it is using:

docker stats

Apparently there’s no easy way to check how much memory pgsql is using when installed natively, interesting lol…

@deltatux Yeah, that's what I've been discovering ... A quick Google didn't seem to help much and while there are a few processes outlined, none of them seem to really help 😛

But as far as I can tell (with the huge caveat that I'm wrong) based on one such Google link, Postgres seems to be taking up only about 300MB. As to whether that's accurate, I have no idea!

@fahim

That's pretty nice if it is indeed 300MB, but might need to be careful as it starts ingesting data, it might start to require more RAM. Though probably with the tuning you did based on pgTune, it probably would cap itself from eating all the available memory, at least hopefully lol 😋.

@deltatux Another post seems to confirm that that's in the right ballpark ... apparently running top and paying attention to the VIRT column is the way to go and that seems to indicate that PG is at at about 360MB ...

So, I guess I give it a day and see how it goes over a longer period of time without me constantly restarting the process to optimize or to check on something 😛