So, I’m still trying out rockstor and today I (hot) removed one drive from a 2 drive single pool. The web-ui stopped responding, and upon checking the display this is what I got (unresponsive):
System is a hp 260 g2 with usb3 disks. last rockstor version (.26, didn’t install the .27 upgrade yet)
(After cold boot and reconnecting the drive the pool is ok again)
Tried the same thing again… this time no problem.
I also updated to .27.
@Zonk Welcome to the Rockstor community.
Note that the single btrfs raid profile, like the raid0 profile, is not a redundant profile. For that your best bet is really raid1 (which has a 2 disk minimum so really 3 disk is a nice practical minimum device count wise). Note that btrfs raid1 will try to ensure that there are 2 copies of all data on 2 independent devices (hence the minumum of 2 disks), even if there are 3 devices, ie it’s 2 way raid1 irrespective of drive count.
http://rockstor.com/docs/pools-btrfs.html#redundancy-profiles
Yes the hot plug scenario is a little touch and go on the btrfs side, although if you could avoid using the usb bus, ie favouring SATA for example, you are likely to get better results as the USB bus is generally not recommended as highly as the SATA bus for such things as disks; especially in multi drive scenarios as it can often do internal resets that affects more than a single device at a time is my understanding. But then not all SATA ports are rated for hot plugging so bit of a catch 22 there. Although I think it was part of the original SATA standard you still get ports / controllers that claim hot plug capability while other don’t.
There is no difference between 3.9.2-26 and 3.9.2-27 that would account for this change in behaviour by the way so this is an underlying technologies type thing: not unless the kernel was upgrade at the same time anyway.
Changes per release version:
Maybe other forum members can chip in here, re the hot plug scenario / kernel panic / USB verses SATA bus etc as it’s not really something I’m that up on myself.
Also note that the missing disk / degraded pool warning in the Web-UI title bar is only updated every 30 seconds:
Hope that helps and thanks for sharing your findings.
Hi. Sure, I know about data redundancy.
I’m just testing how the system handles various things so I know what I can trust to use for my data backups. I don’t expect an instant fix for this issue, but I still think it is important to point it out.
I understand that just removing a drive is not something you should do - it is something that can happen though, especially with usb drives. Kernel panic from this seems a bit extreme.
I like rockstor, but I will report the various issues that I find - because there is still much to improve.
Most welcome and totally agree.
Re the kernel panic: we use an unmodified elrepo kernel-ml package. There are however plans afoot, and ongoing efforts, to improve on our current piecemeal construction; as always everything takes a lot longer than one would hope.
Hang in there and keep the observations / reports coming. I think Rockstor’s code, and the technologies it tries to democratise (and depends upon) are getting better all the time, if only in small steps mostly.
