Adding storage on the fly

#1

If your system supports hot swapping drives, are you able to add storage to your 3node without having to take it down and reboot the OS? Will the OS autodetect a new unallocated drive while running?

Thanks!

#2

From the experience of previous farmers, you should reboot the 3node when you hot swap drives.

Good question btw.

#3

When I did this with SSD’s my dell R720 started reading the SSD as HDD’s

1 Like
#4

I also tried this on an HP dl360p and the ssd was not recognised at all until I rebooted the server. But to be honest I had no patience and only waited a couple of hours

1 Like
#5

Yeah I waited a full 24 hrs and nothing updated on the portal.grid or the explorer. So I rebooted and
the new drive showed up. Should the portal be updating our resouces? I kept expecting to see the HRU change but it didn’t. Does the hourly check in just measure uptime?

#6

Its a good question. Someone should try to see if removing storage and booting in another node is a way to cheat. Infinite storage hack. Little things like that need to be fixed up if it actually works.

1 Like
#7

Excellent point.

Obviously the easiest “hack” as of now is to revert every v3 node to v2 node to get more TFT.

What you stated here would be a big vulnerability. We should check this.

#8

Wow… didn’t even thought that far. But you are right… we have to make sure that this would not be possible!

1 Like
#9

Yeaaah you’ll want to tighten this up. Its not quite infinite storage hack, you are still constrained by storage on hand. There’s no telling how prevalent it might be but this can’t have been the first time its been thought of. Seems like a pretty easy fix is for the hourly ‘node check-in’ to just check and update the state of each resource in addition to logging the uptime.

Things change without maliciousness as well. A hard drive dies, a server power supply goes out forcing the Xeon’s to cut cores to conserve power on just the backup, etc. Having it update the state of your machine would actually be helpful in a management sense to tell when things other than “the node is just fully down” have occurred.

#10

@sigzag, we’re the drives that detected as hdd in the dells both Samsung and behind a raid controller?

#11

I tried hot swap disk on a HP gen 8 behind a p410 raid controller with intel enterprise/datacenter SSDs.

Had to reboot to get them recognized

#12

Interesting, I’ve been doing a lot of testing on the p410 and I think this is actually more related to it than zos, the controller doesn’t clear the information from the old drive until a new drive replaces it,

In the case Of Samsung drives they didn’t fill all the blanks and data from the old drives ended up staying on the slot with the Samsung drive in place.

As long as the drive isn’t a Samsung consumer drive or something similar a reboot seems to clear everything up, in the case of the Samsungs I had to wipe the controller and make that drive the only one the controller has seen in that slot

#13

Word from @azmy in this issue is that adding disks while a node is running is not supported. Removing them, on the other hand, should be detected and reported capacity adjusted accordingly. Did anyone actual demonstrate that removing disks did not cause numbers in the explorer to adjust accordingly?