Upgrade HDDs in NAS542

Joe

Hello,

I have a Zyxel NAS542 with 4 x 3Tb WD Red HDDs in RAID5. So this is the maximum number of drives that can be installed. Now I got the issue that the NAS is almost full and I need to upgrade the drives.

So I got 3 x WD RED 8Tb HDDs to replace the old 3Tb drives. So I should get 16Tb in RAID5.

Now I am asking myself if I just can replace a 3Tb drive to a 8Tb one and wait until it got rebuild. 
Then replace the next and then the last of the three drives. 

Is it then possible to pull the last 3Tb drive out and tell the NAS to use just the three 8Tb drives in RAID5?

Is this possible without the need to set the NAS up from scratch?

#NAS_December

Mijzelf

I'll add some background. The NSA5xx runs a strange CPU (a Mindspeed Comcerto), which can use both 4kbyte or 64kbyte memory pages.
On fw 5.00 ZyXEL choosed to use 64k memory pages. In theory this is faster, as less memory administration overhead is needed. 
And another advantage, for some technical reason a cluster in an ext3 filesystem has to fit in a memory page. And ext3 supports at maximum 2^32 clusters. Which means that using a 64kb page size gives the advantage to create a 256TiB volume. So on fw 5.00 the volume size was not limited for all practical purposes.
A big disadvantage is that such a volume can't be mounted on any CPU using smaller memory pages. (And virtually any other CPU uses 4k pages). So your precious data is in danger if the NAS fails. (OK, there is a trick to solve that).
There a other disadvantages. So you can't run any software which is not specifically compiled for 64k pages. Which also means the firmware for the NAS326, which runs another CPU, could not be identical. And maybe there are other disadvantages I'm not aware of.

So on fw 5.10 ZyXEL decided to swap to 4k pages. AFAIK this is actually a bit faster, although 64k in theory can be faster. Maybe because the kernel (and samba!) is optimized for 4k pages.
Anyway, this gave (AFAIK) one disadvantage, a volume could not be bigger than 16TiB. 
Boxes with a volume >16TiB cannot be upgraded, without loosing the possibility to mount that volume. The work around is to first shrink the volume. 
The free space can be used to create a 2nd volume.

In theory the used ext4 filesystem can handle more than 2^32 clusters. But the SDK Mindspeed provides for this CPU doesn't support that.

Fredzoul1

@Joe
You construct your RAID5 with 4 HDD, now you can't reduce the number of HDD for the RAID without save data in another place and delete the volume for make a new one.
You have to replace one HDD, wait until the RAID is rebuild and then change the second, wait rebuild and change the third. But you can't have most place on your RAID without changing the last HDD, because the free space is calculated from the smallest HDD. This is what you can do with GUI. But, here in the community, exist a person named Mijzelf, and he have many advanced knowledges and tips, perhaps he find another solution to help you :-)

Joe

Thanks for the answer!
So it would be possible to change one, then rebuild and then change the next if I do this for all 4 HDDs?
And then when all drives are changed to 8Tb ones is it possible to use 24Tb RAID5?

Ijnrsi

@Joe, 
Yes it is possible to expand compacity from 3TB to 8TB, but you need replace and wait for first rebuild doen to replace second one.
NAS542 is using 4k page size for the volume in ext4 format, the maximum capacity per volume is 16TB, so the forth HDD you need create as a single volume.
https://ext4.wiki.kernel.org/index.php/Ext4_Disk_Layout

Fredzoul1

The max capacity seem to be up to 24 TB

Ijnrsi

@Fredzoul1
Yes, the max capacity can up to 24TB, but not volume.
NAS540 has limitation since firmware:5.10(AATB.1)C0 as below:


NAS542 is newer than NAS540 and released after v5.10(AATB.1)C0, so NAS542 should also have the same limitation.

Fredzoul1

@Ijnrsi,
Thanks you very much for this precision, I don't know this detail before, useful to know this for the futur :-)

Mijzelf

I'll add some background. The NSA5xx runs a strange CPU (a Mindspeed Comcerto), which can use both 4kbyte or 64kbyte memory pages.
On fw 5.00 ZyXEL choosed to use 64k memory pages. In theory this is faster, as less memory administration overhead is needed. 
And another advantage, for some technical reason a cluster in an ext3 filesystem has to fit in a memory page. And ext3 supports at maximum 2^32 clusters. Which means that using a 64kb page size gives the advantage to create a 256TiB volume. So on fw 5.00 the volume size was not limited for all practical purposes.
A big disadvantage is that such a volume can't be mounted on any CPU using smaller memory pages. (And virtually any other CPU uses 4k pages). So your precious data is in danger if the NAS fails. (OK, there is a trick to solve that).
There a other disadvantages. So you can't run any software which is not specifically compiled for 64k pages. Which also means the firmware for the NAS326, which runs another CPU, could not be identical. And maybe there are other disadvantages I'm not aware of.

So on fw 5.10 ZyXEL decided to swap to 4k pages. AFAIK this is actually a bit faster, although 64k in theory can be faster. Maybe because the kernel (and samba!) is optimized for 4k pages.
Anyway, this gave (AFAIK) one disadvantage, a volume could not be bigger than 16TiB. 
Boxes with a volume >16TiB cannot be upgraded, without loosing the possibility to mount that volume. The work around is to first shrink the volume. 
The free space can be used to create a 2nd volume.

In theory the used ext4 filesystem can handle more than 2^32 clusters. But the SDK Mindspeed provides for this CPU doesn't support that.

Joe

Thanks for all these informations. Now I know for sure that I am not able to create a single Volume bigger than 16Tb.
So I will cerate two Raid 1 with 8Tb each