The product information for EDS references “RouterOS v7 (Special ROSE edition)”. Is it regular RouterOS with the rose package on top, or is there really a special edition?
ROSE Storage module is a nice thing, but advertising THIS appliance with BTRFS and without (open)ZFS support is rather a no go atm…my 2 cents.
Hope ROSE will evolve into supporting ZFS over time…
With containers and btrfs, you just use subvolumes. No need for overlayfs. Not sure about the state of docker storage driver for zfs, but the btrfs one works well.
However, with btrfs, I would be interested how the parity story for the underlying storage was solved. Probably not at all. Also, how to ensure HA? Imagine you need to update this device and there’s a bunch of NVMe-TCP, iSCSI and NFS clients running.
Interesting, Chateau LTE12 (2025) now with 32MB flash memory and different modem (EG120K-EA). Is this the end of 16MB era? The leaked “7.20_ab28” packages in 7.18 topic already shows about 700kb larger routeros-package, which can’t fit routeros+wifi-qcom-ac into 16MB flash devices anymore. Hope the best.
Regarding BTRFS, I strongly advise against using advanced RAID setups or filling up the storage. These factors, combined with other known issues and the lack of clear documentation on storage configuration, pose a significant risk.
If a filesystem or storage failure occurs without proper repair tools, it can become a serious problem. Mikrotik should include a warning about this, as other Linux distributions do. Additionally, they should provide detailed instructions and the necessary tools for handling filesystem and storage failures.
It’s quite obvious that storage systems are not Mikrotik’s area of expertise.
A device that is recommended instead of the cloud that uses proprietary and closed software is worse and provides fewer guarantees.
Here in fact comes the problem of using “in house” solutions rather than the cloud.
If you do everything yourself and something breaks you get screwed, if you don’t spend triple and do everything redundant and with an additional separate backup.
Exactly why I was proposing ZFS.
I am running several local storage appliances / NAS for Storage and Containers and had nothing but uncomfortable experiences with BTRFS, especially in RAID configurations and high -performance Dockers, like influxDB.
Since moving to openZFS all this is stable as rock and good ZFS documentation & training is also available in abundance.
ZFS is the de-facto enterprise storage filesystem standard and available on recent linux kernels with openZFS. I can only hope that MT can move to openZFS, given the fact that they might not use the recent linux kernels usable with openZFS.
What do you understand under the phrase “advanced RAID setups”? Please, no hand waving.
RAID can do either mirroring (raid 1), stripping (raid 0), combination of both (raid 10), or parity with one redundant drive (raid 5) or two redundant drives (raid 6). Btrfs can do 1, 0, 10 perfectly fine; with 5 and 6 it has write hole.
The write hole may or may be not an issue for you. You can solve it by using an UPS or battery backup, if you really want to. Or do the Synology approach (btrfs on top of dmraid for parity). However, 5 and 6 are not used in enterprise settings, since it is more efficient to just use some form of raid 10. Raid 5 and 6 are enthusiast / smb forte. And since Mikrotik aims at enthusiasts and smbs, that’s why I asked about the parity.
EDIT: zfs technically has no raid5 or raid6. It has raidz and raidz2 (and raidz3). It has slightly different semantic than raid5 and raid6, and by this redefinition, it can close the write hole. But make no mistake – it is not raid5 or raid6.
If there is any problem with filling up the storage, it is due to lack of understanding how allocating blocks in CoW filesystems works. In all the years people running Synology boxes have yet to have problem with it.
ZFS is a legal problem. There’s a reason why it is an out of kernel module, and it is basically undistributable. For now, only Canonical, Proxmox and TrueNAS took the risk. They seems to be small to warrant attention; but otherwise, people asking for uncomfortable things might knock on your door.
ZFS is not “de-facto enterprise storage filesystem standard”. The enterprise NAS-es are ruled by proprietary filesystems, which are more advanced than ZFS. They are also patented, and some of those patents could be threat to Oracle… and that’s about one reason, why ZFS is GPL2-incompatible, so Oracle layers do not have to defend their company, but can spend their time on much more profitable attacks on their clients.
From technical side, ZFS had also pretty serious limitations, that its promoters were quiet about. Like not able to replace all your drives with bigger ones and expand your pool (“just add the drives and create a new vdev”). Or that nasty bug recently where ZFS was corrupting data.
Also, what do you mean with high-performance Dockers? The btrfs storage plugin for docker and k8s is in way better shape than the zfs one. Probably because it is used in the real world; Meta runs their containers on btrfs. Who uses zfs at that scale?
So its is not that black and white, as you paint it. Zfs has its advantages, but also disadvantages. Btrfs has different advantages and different disadvantages.