Are drives hot swappable without a BackPlane?

level42

New Member
Feb 22, 2017
13
1
3
32
So, I see chassis that are often advertised as "Hot Swappable" which, typically means they have a backplane of sorts.

My question is, are drive not connected to a back plane hot swappable?

For example, if I have drive connected to my Perc H310, using SSH 8087 to SATA breakout cables, would they technically be hot swappable? Or is this a bad idea, and I should invest in a backplane of sorts?
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,029
1,578
113
CA
"Hot Swappable" usually means you can access the drive and remove it without turning the server on or off, not so much that it has a backplane. Removing the chassis, lid, and to do that you should power down or it will throw errors, run fans at 100%, and generally not a good idea and thus not hot swappable are the drives 'inside' the chassis.

I disconnect \ connect drives all the time with HBA, onboard Sata, and even in windows (with varying luck) and it works just fine.
 

level42

New Member
Feb 22, 2017
13
1
3
32
Ok, so I assume as long as I'm careful, and ignore any chassis intrusion errors, it'd be safe to disconnect a drive on the fly?

If I use a system like FreeNAS, and it reports a dead drive, is it safe to power down the server with a dead drive? or would I need to rebuild the raid before a power cycle?
 

seang86s

Member
Feb 19, 2013
144
12
18
Same here. I've "hot plugged/unplugged" SATA drives too without a problem. Don't wiggle the connectors, pull straight out and push straight in. On some consumer boards, you need to mark the SATA port as hot swappable in the BIOS or they won't recognize the change.
 

level42

New Member
Feb 22, 2017
13
1
3
32
Same here. I've "hot plugged/unplugged" SATA drives too without a problem. Don't wiggle the connectors, pull straight out and push straight in. On some consumer boards, you need to mark the SATA port as hot swappable in the BIOS or they won't recognize the change.
Thanks for the additional tip.
 

ttabbal

Active Member
Mar 10, 2016
747
202
43
43
Ok, so I assume as long as I'm careful, and ignore any chassis intrusion errors, it'd be safe to disconnect a drive on the fly?

If I use a system like FreeNAS, and it reports a dead drive, is it safe to power down the server with a dead drive? or would I need to rebuild the raid before a power cycle?

Shutting down to replace a drive is fine. I always try to get the serial number and verify when pulling the drive to double check I have the correct unit, even in a backplane chassis. I just copy/paste it from whatever machine I'm using to manage it. When you reboot, ZFS will detect that the drive is different and not part of the array and show your array as degraded. Do a zpool replace and it will add the new drive and start copying the needed data over for you.

You can disconnect on the fly with SATA and SAS. Pull data first, and try not to bump any other connectors. That last bit is easier said than done with most internal mounted drives. They tend to be packed closely and pulling the wrong connector could fault the array. Not the end of the world as ZFS will generally figure it out, but not a great idea either.
 

Fritz

Well-Known Member
Apr 6, 2015
2,120
483
83
66
I learned the hard way not to monkey around inside a chassis while it's running. You might let the magic smoke out and then it be dead. :(
 

acquacow

Well-Known Member
Feb 15, 2017
578
299
63
38
Everything connected to PCIe is pretty much hot-swappable too. The trick is to inform the OS to unload the drivers before you do it.

I've hot-swapped my GPUs before, just right click in device manager, disable the device, then once it's unloaded, pull it out of the slot, unplug the power cables, plug in the new GPU power and install the card.

Windows will re-detect it and bring up the display.
 

pricklypunter

Well-Known Member
Nov 10, 2015
1,606
470
83
Canada
I would advise caution here. Backplane's that are designed for hot-swap have power management controllers on them, such that the power supply is effectively decoupled/ isolated from the load and unload effects presented by sudden removal or application of power to a disk. Even if the data plane is capable of hot-swap, just the inrush current spike from a disk can upset the apple cart and cause a PSU to foldback for a split second, taking down the whole server suddenly and without warning. Have I done it myself in an emergency? Yea sure, but I weighed up the risks of doing so first and made sure my back-ups were up to date. I would never do this on an enterprise host in production without having permission signed in blood by the CEO, or advise anyone else to do this unless you understand what might happen and are prepared for the outcome should it all go sideways. This sort of thing is really only for the homelab and only when you have no other option available to you. I would definitely limit the number of times you try this, being successful a handful of times doesn't make it a reliable method, you might just pop the drive controller or interface next time.
 
  • Like
Reactions: Aestr and i386

EffrafaxOfWug

Radioactive Member
Feb 12, 2015
1,242
420
83
Just to expand a little on what others have said - it should work as long as the SATA/BIOS is set to hotplug; lots of motherboards from a few years ago defaulted to disabling hotplug for whatever reasons.

The SATA power plug itself has contacts of different lengths in order to disconnect the power in as clean a fashion as possible, but if you've got a marginal (or just plan crap) PSU then I've seen it cause power issues before, and while windows' handling of disappearing drives is usually OK (I've not seen a drive hotplug hang in years), other software might still throw a tizz.

Although personally the value of having hot-swap slots in convenience alone makes them well worth paying for, especially after you've had the tops of your knuckles sliced off by fans one too many times ;) By far the most common problem I've run into now when working inside a case whilst it's running (it's happened to me three times now) is accidentally shorting something out; once a dropped screw that bridged two capacitors (locked up the system), once a hard drive whose cover brushed against the underside of an HBA when coming out of the internal bays (temporarily took out one of the SAS ports resulting in four hard drives dropping immediately out of the array), once a slip of a screwdriver that jogged a reset.

Get a hotswap case with a backplane if your means allow it :)