So far my experience with Nextcloud has been that it is a pain in the arse to install, and once it’s installed is slow as anything. Literally couldn’t run it on my pi 3b, now got it up and running pretty nicely on a NUC but it’s still not great. Have caching set up.
I have the notes app installed on my android phone and I can never used rich text editing because it gives timeout error.
This shouldn’t be this complicated. All I want is to de-Google my documents and notes, and self-host my kanban. I don’t really need the rest though it’s nice to have the options.
Do people use alternatives? Am I doing something completely wrong? I set it up using nginx which I know is not supported, but the alternative using Docker AIO didn’t allow me to use custom port easily.
I suspect nextcloud having performance issues with slow Disk IO. With rootless containers I had a much worse performance than rootfull. Also using MySQL Backend instead of SQLite did speedup the performance.
Nevertheless I have the same problems with nextcloud as you stated. Pretty much not as usable as I thought.
It’s on a SATA drive, albeit hard drive not ssd and I’m using mariadb. Everybody seems to suggest I need a beefier server but as a developer myself, the functionality of the software doesn’t seem to warrant anything more powerful.
How much memory? I think nextcloud wants around 8gb to run happily (ymmv). I’ve tried it with smaller sizes and ran into issues.
Yes I have 8gb of ram, but it seems insane that it needs that much considering what it is doing.
I think nextcloud wants around 8gb to run happily (ymmv).
As a developer myself, where did it go wrong?
I’m not sure it’s going “wrong”. It depends on the scenarios it’s designed for. If they intend it to be run on servers (there is no class of raspberry pi that is a server) then you design it to take advantage of those resources.
But it’s designed for selfhosting and consumes all that without doing much.
I didn’t mean to imply it needed server hardware. You can absolutely self-host Nextcloud. But RPIs are the absolutely lowest-end of hardware for serving duties. They’re great little systems but they’re designed to be cheap, not performant.
I self-host currently on a VM running on a 12 year old x86 system with 8GB RAM and with the Nextcloud file storage going over a 1Gbps NFS mount. Not exactly a high-end setup. And it performs just fine. I was previously running on a AWS EC2 instance where I noticed occasional issues running on a T4g.SMALL. (only 2GB RAM). I had to bump up to a MEDIUM at some point though.
It worked with less RAM pretty fine for a long time. But as I increased usage it would have issues occasionally. I think with all the images I have it was doing lots of processing for thumbnails and the like. I never really dove into it to see what exactly was going on though…
But still - a moderately old desktop system with 4-8G of RAM is just fine for “self-hosting”.
EDIT: I should add - I’m also hosting MariaDB on the same server - also with its data stored on an NFS share.
I have nextcloud running on docker on a Raspberry Pi 4 and I’d say the performance is comparable to Onedrive web interface. If you’re getting timeouts then something must be wrong with the setup, not the machine it’s running on. Using Postgres instead of MySQL or using an SSD instead of HDD is not going to help your issue.
I’m my experience, nextcloud is quite I/O bound. The performance of your storage device will greatly affect nextcloud performance. But if you’re already using SSD and the performance still bad, maybe there are other issues with your setup.
For me, speed isn’t the only issue. Everything about nc seems to be cobbled together in the most inconvenient way possible. Updates have always been hit or miss for me - and if you choose to use dockerized versions, you might as well shoot yourself, everything is very slow, even as the only use having it running on a quite capable machine it feels sluggish (not slow, but uncomfortably delayed).
It’s a glorified Dropbox clone, why do I need anything more than a rpi1 for that?
Just want to say that I’ve been there. There was a time my Nextcloud install was incredibly slow. Fortunately (or unfortunately?), it is featureful enough and widely supported that once you figure this issue out, it is a nice service to keep running.
For me, adding Redis was essential. It doesn’t really make sense to me why (nothing I do on Nextcloud is intensive or data heavy) but it has greatly improved the performance of my app.
My entire setup is a containerized Nextcloud, Nextcloud Cron, MariaDB (if I knew Postgres was an option, I would’ve chosen that), and Redis:
version: '2' services: nextcloud: container_name: nextcloud image: nextcloud:27-apache restart: unless-stopped environment: - MYSQL_PASSWORD=nextcloud - MYSQL_HOST=db - MYSQL_DATABASE=nextcloud - MYSQL_USER=nextcloud labels: - 'public-service=true' - 'traefik.enable=true' - 'traefik.http.routers.cloud.rule=Host(`nextcloud.some.domain`)' - 'traefik.http.routers.cloud.tls=true' - 'traefik.http.services.cloud.loadbalancer.server.port=80' volumes: - /some/data/dir/nextcloud/data:/var/www/html - /some/external/dir:/wew:ro nextcloud-cron: image: nextcloud:27-apache restart: unless-stopped command: [/cron.sh] environment: - MYSQL_HOST=db - MYSQL_DATABASE=nextcloud - MYSQL_USER=nextcloud - MYSQL_PASSWORD=nextcloud volumes: - /some/data/dir/nextcloud/data:/var/www/html - /some/external/dir/:/wew:ro db: image: mariadb:10.4 restart: unless-stopped environment: MYSQL_DATABASE: nextcloud MYSQL_USER: nextcloud MYSQL_ROOT_PASSWORD: nextcloud volumes: - /some/data/dir/nextcloud/db:/var/lib/mysql mysqldump: image: mariadb:10.4 depends_on: [db] # restart: never # cronjob labels: - 'cron.schedule=0 0 8 * * ?' entrypoint: [mysqldump, -h, db, -u, nextcloud, -pnextcloud, --all-databases, -r, /out/nextcloud.sql] user: root volumes: - /some/data/dir/nextcloud/db-dump:/out redis: image: redis restart: unless-stoppedCan I ask why the separate NC container for cron? Also, I presume the mysqldump container is for easy db backups?
The separate cron container made the most sense to me. Other variants “work” but imo are mostly workarounds to avoid setting up a real cronjob. Beyond this I have no real reason, nor can I vouch that is is more or less performant than others.
Yes, the mysqldump container is for easier restores. It’s much easier to restore from a .sql file than a raw data dir that was copied while the DB was running ;) (speaking from experience…)
