Over the last two weeks, I tried out the RAIDZ expansion feature with an 18TB drive in my main pool. I encountered a few issues that could be handled better. When performing a RAIDZ expansion, the current process can be optimized to improve user information and system performance.
Estimated Duration Information:
Clearly inform the user of the estimated duration for the RAIDZ expansion process. This can help set realistic expectations and plan accordingly(~50MB/s,https://openzfs.org/w/images/5/5e/RAIDZ_Expansion_2023.pdf).
Automatic Deactivation of Scrubs:
During the expansion process, automatically deactivate scrubs for the affected pool. This is to prevent significant IO WAIT issues, which can severely impact system performance. Once the expansion is completed, scrubs can be reactivated.
Progress Display in Task Manager:
Integrate detailed progress information from zpool status into the Task Manager. This includes the current data transfer rate, percentage completed, and estimated time to completion. For example:
expand: expansion of raidz1-0 in progress since Wed Jun 19 01:12:14 2024
1.99T / 50.9T copied at 54.8M/s, 3.91% done, 10 days 19:48:35 to go
I agree, i couldn’t even find the option, i kept trying from expand and other entry points like the ‘you have 1 disk unused tile’ agree with all your points.
Actually, most/all of the data in the vdev gets shuffled around. It stays at the same offset within the vdev, but it gets shuffled around to match the new geometry.
It preserves existing parity - So it keeps the data
Allows for online expansion - so reads and writes can continue immediately
But it shuffles the data to free the space - that takes a while.
It took ~ 11 days (~55M/s). Raidz 5 pool was filled at 75%, 4x 18TB. Now 5x 18TB. I’m on the nighty train. It would have taken longer If I hadn’t stopped the scrubbing. In the nedata screenshot you can see right what happend after I stoppend the scrubbing.
The stable build came out today. Unfortunatly, the user experience has not been improved. I have the feeling that TrueNAS is working on expanding my 4x 4 TB to 5x4TB Z1 pool, but the interface got stuck at the same screen as the OP.
Wonder if there is a way to check the intermediate progress manually?
I am running a RAIDz expansion (says it’s going to take 10 days!) and I also had poor UI experiences.
First, the job sat all day at 25%. It happened that I needed to restart the server, and since I had no idea how long this was going to take I just bit the bullet and did it.
When it came back the drive showed as unavailable. Ok, my server is kind of weird, I can accept that.
Upon rebooting again, all drives were visible. The pool showed the same capacity as before, but with all 5 drives. There was no indication that the expansion was still happening.
I confirmed via the shell that the expansion is in fact still happening, though it is going to take a very long time >_>
I’m a TrueNAS newb coming over from a LONG time Synology user. My build has been in progress. Today I added a 2TB drive to my existing 4 drive zvol. After an hour I get the following:
pool: p1
state: ONLINE
expand: expansion of raidz1-0 in progress since Tue Mar 4 18:38:45 2025
373G / 1.26T copied at 124M/s, 28.90% done, 02:06:43 to go
config:
NAME STATE READ WRITE CKSUM
p1 ONLINE 0 0 0
raidz1-0 ONLINE 0 0 0
3ea502fa-c4d0-4c55-a33f-3c865b0b8a68 ONLINE 0 0 0
f439046e-5a86-402c-b098-ef71721c0101 ONLINE 0 0 0
ad7dc358-6e22-4f70-98d1-f4134fb850b6 ONLINE 0 0 0
fb73c4f5-7531-41e4-9983-d738a42f5bea ONLINE 0 0 0
e52390cc-a664-4b22-b27d-5d9ffd0334c9 ONLINE 0 0 0
logs
9f9c3599-40c2-4c0f-b4ca-5b23a8ba8dcc ONLINE 0 0 0
cache
0a9dd1f0-b579-4c40-81f5-bb88ed209e07 ONLINE 0 0 0
errors: No known data errors
I agree with most of the comments about improving the UI. TrueNAS can use a lot of improvement. Most operations have been painful or confusing coming over from DSM. The virtualization is especially troublesome. Either the shares on this zvol are offline during expansion, or my first attempt add creating one samba share and one Time Machine share have failed miserably. They are not connecting from my Mac (mini on latest Sequoia). What I will say on virtualization is my Synology stopped working with any Redhat newer than version 8. Seemingly because they dropped support in the kernel for the old Atom processor. TrueNAS on my 3 year old i7/64Gb IS running Rocky9, after a LOT of fiddling with settings to find a combination that worked.
Your p1 pool will lose all its data if your LOGS device goes bad. You usually want the LOGS to have at least the redundancy of the main pool. You should have a mirror pair.
EDIT
I thought the LOGS listing was not SLOG.
You also should check your hard drives to make sure they are CMR and not SMR.
BASICS
iX Systems pool layout whitepaper
Special VDEV (sVDEV) Planning, Sizing, and Considerations
What? Neither SLOG nor L2ARC is essential to the pool; the data will be just fine (with the exception of anything in flight during an unexpected loss of power) with failure of either or both of them. There’s not likely to be any good reason for SLOG with a parity RAID pool, but it still isn’t dangerous.
There’s no need for SLOG to be redundant. Data is still written from RAM, not from the copy in SLOG. If the NAS has an unclean shutdown AND the SLOG fails on reboot, then you possibly lose data from the last txgs—pool can still be imported by discarding transactions and/or using -m. If the SLOG fails but the NAS keeps running, ZFS will revert to its default ZIL: No data is lost, just some performance.
And the ZIL is a non-redundant stripe across pool drives.
The actual question is whether a raidz1 of smallish drives needs a SLOG and L2ARC to begin with… I do not know of a workload which requires sync writes and for which raidz1 is an appropriate geometry.
