(Older) server hardware (e.g. Dell 720d)- what do I need to know

[Whiskey]

I'm reviewing options to replace my aging NAS, and found that old servers with lots of RAM are not that expensive. Specifically I found quite a few Dell R720d being sold, that look like a good fit for what I need.
However, I haven't worked with servers yet and while I'm fairly competent with Linux and (to a lesser extent) FreeBSD, I'm sure there will be some things to be aware off. I already came across posts that said not to use the pre-configured RAID controller, but to get a proper HBA instead (I assume you need to disable the RAID controller then). What else is there to watch out for? Any experience (did this work well for you?), comments or pointers to other documentation or threads is much appreciated. Thanks!

 

[firesyde424]

We have several Dell PowerEdge R720xd servers operating as TrueNAS core head units. As with most things, servers become less reliable the older they get, so that needs to be taken into account. In our case, these TrueNAS servers have been moved to non-production roles, in most cases becoming backup storage. We have had issues with the Broadcom NICs that originally shipped as OCP cards on a those servers and have either had to disable them or completely remove them. Another issue we had was the on-board RAID controller that shipped attached to the backplane. The PERC 710, I believe. While it worked with TrueNAS and FreeNAS before, we had long term hot swap issues that effectively meant we couldn't use the R720xd chassis slots for capacity storage instead opting to only use external JBODs.

One final issue we experienced with the R720xd that was never addressed has to do with Dell's BIOS watchdog service. At random intervals, and for no consistent reason we were able to divine, Dell's watchdog service would "detect" that TrueNAS had hung and reboot the server. In reality, the server was perfectly fine and was not experiencing any issues which meant that watchdog caused an outage for no reason. We ended up disabling the watchdog service via the BIOS setting for all our R720xd and R730xd servers running TrueNAS.

 

[Whiskey]

Thanks, this is super useful.
Can you elaborate a little on the RAID problems? Would you normally be able to remove the RAID controller and plug in an HBA, but you couldn't with that model? Still am completely new to this I'm afraid.

 

[firesyde424]

Thanks, this is super useful.
Can you elaborate a little on the RAID problems? Would you normally be able to remove the RAID controller and plug in an HBA, but you couldn't with that model? Still am completely new to this I'm afraid.

We inherited the R720xd servers from an IT services company via the purchase clause of a leasing agreement and did not have control of their configuration. The issue had to do with the PERC H310 that was installed when the IT services company originally purchased those servers. I don't remember the exact circumstances as this was some years ago. Whatever the cause, it effectively meant that the 12 x 3.5" drive bays on the front of the servers could not be hot swapped. The servers had to be shut down to replace a failed drive in those slots. The issue was not present if the drives were being used in a RAID controlled by the H310, only if the drives were in passthrough mode.

You are correct that the internal PERC H310 could be replaced by an appropriate HBA. However, in our case, the effort in replacing the cards is not worth the minor increase in storage capacity that would be realized. These servers are very much in a "if it's not broke, don't fix it" mode. Several of them were perfectly happy to run FreeNAS 9.11 with 3+ years of uptime. The only reason they were upgraded to TrueNAS at all was because the legacy versions of FreeNAS could not receive snapshots from the newer TrueNAS builds, which effectively negated their use as backup targets.

 

[Jessep]

Dell R720 are fairly old (10 years) and will have higher power consumption and shorter useful life.

I would suggest R730 with V4 CPUs as the current sweet spot (or V3 CPUs and plan to upgrade when they are cheap). There will be a larger upfront cost.

 

[Whiskey]

Thanks, after researching further last night that notion came to my mind as well. I expect the next build to last at least 5y as well (hopefully closer to 10) so a bit higher cost up front seems worth it.
Had it on the list to look into it a bit more tonight, e.g. see if that raid controller can also be flashed.
You can upgrade the CPUs without problem? That's good to know

 

[firesyde424]

Dell R720 are fairly old (10 years) and will have higher power consumption and shorter useful life.

I would suggest R730 with V4 CPUs as the current sweet spot (or V3 CPUs and plan to upgrade when they are cheap). There will be a larger upfront cost.

This is pretty sound advice. These are now 2 generations old but some relatively young examples can still be found for reasonably cheap as they have been age out of enterprise infrastructure.

 

[Whiskey]

I'm going to look around a bit. So far these (and same for 630 vs 620) are quite a bit more, because of the DDR4 I think. But thanks, doing the research is half the fun ;-)

 

[Daisuke]

I purchased a Dell R720xd recently, will keep it for at least a decade. I rather spend the saved money for quality disks (which are way more important to last), I use 12 HGST Ultrastar He8 Helium disks for default pool and 2 SSDs for services pool. In contrast, I had a Dell C2100 for a decade with zero issues. Still going strong and super silent, like the R720xd (with the right firmware combination). The only reason I upgraded is to transform the C2100 into a backup for important stuff and keep it powered down. Dell have super robust hardware, never had a single issue with their servers, running 24/7.

My C2100: