Ode to the Dell C2100/FS12-TY

[Jason Hamilton]

So I'm wondering if I would benefit from slapping a SSD into this box to use as L2ARC since I'm currently running with the 96GB of ram in it. My ARC hits are hanging out around 70% and the ARC itself is up at 84GB. I've read so many things that go back and forth on the L2ARC. But after reading @RickH talk about his and the fact that I do have a few VMs that are running off of my nas. Granted that this is just my home use box and I have fun experimenting with things with it. I can tell you this much going from 16GB in the old build up to the 96GB in the new I do see a performance difference there with it.

 

[RickH]

So I'm wondering if I would benefit from slapping a SSD into this box to use as L2ARC since I'm currently running with the 96GB of ram in it. My ARC hits are hanging out around 70% and the ARC itself is up at 84GB. I've read so many things that go back and forth on the L2ARC. But after reading @RickH talk about his and the fact that I do have a few VMs that are running off of my nas. Granted that this is just my home use box and I have fun experimenting with things with it. I can tell you this much going from 16GB in the old build up to the 96GB in the new I do see a performance difference there with it.

If you have any RAM slots left you're going to benefit more from adding RAM than L2ARC. If you have multiple VM's running with the right workload you may see some very slight performance gains, although they would be pretty hard to quantify in a home-lab.

As an example I have a storage server currently backing 25 vm's with the following ARC stats:
ARC: 113.9 G - 64..3% Hit Ratio
L2ARC 199.7 G - 45.3 % Hit Ratio

I have another server configured pretty much the same way, but it's used for SMB archival shares instead of VM hosting:
ARC: 110.4 G - 13.3% Hit Ratio
L2ARC 187.0 G - 0.2% Hit Ratio​

Obviously the money I spent on the SSD in the 2nd server was pretty much wasted and I should probably re-evaluate the config (I'm just too lazy to open it up to pull it out at this point).

It's ultimately up to you how you spend your money, but I'm guessing that you're not going to notice much of a difference by adding a L2ARC device...

 

[Kevin Horton]

As an example I have a storage server currently backing 25 vm's with the following ARC stats:
ARC: 113.9 G - 64..3% Hit Ratio
L2ARC 199.7 G - 45.3 % Hit Ratio

I have another server configured pretty much the same way, but it's used for SMB archival shares instead of VM hosting:
ARC: 110.4 G - 13.3% Hit Ratio
L2ARC 187.0 G - 0.2% Hit Ratio​

Obviously the money I spent on the SSD in the 2nd server was pretty much wasted and I should probably re-evaluate the config (I'm just too lazy to open it up to pull it out at this point).

Which version of FreeNAS are those two servers running? The reason I ask is that we now know that arc hit ratios prior to FreeNAS 9.10.2 were artificially too high. There was a bug that caused that accessed the same file(s) very frequently, which meant those files always had hits from the ARC. 9.10.2 fixed that bug, so the reported hit ratios are now much lower (and correct).

 

[RickH]

Which version of FreeNAS are those two servers running?

The first one with the higher hit ratio is on:
FreeNAS-9.10.2 (a476f16)
Looks like the last upgrade/reboot was 28 days ago.

The 2nd archival one with the lower ratio is running:
FreeNAS-9.10-STABLE-201606270534 (dd17351)
I haven't upgraded/rebooted this one in a while - currently at 186 days uptime​

I wasn't aware of the artificially inflated ARC stats on the older version - but the server reporting the high numbers is running 9.10.2...

EDIT: I just looked at the graph in the GUI and it looks like there's still some type of ARC reporting issue - the graph is reporting that I have 1.5T of L2ARC - which is pretty impressive given the fact that the arc is housed on a 256GB Samsung 850 Pro:

[

 

[Jason Hamilton]

Thank you @RickH for your input there. I do still have more ram slots available and I have more 8GB ddr3 sticks that I can slap in there. I guess I'll max it out.

 

[St KaJa]

Thanks for all the information! As a result of this thread, I ordered a c2100 from dfwpartsdepot.com and look forward to setting it up this weekend. I also bit the bullet and ordered a JPYNN 10GB mezzanine card from eBay to avoid using one the pcie slots. I haven't confirmed it is working yet, but the controller is marked (under the heatsink) as a JL82599ES and seems quite capable from the Intel datasheet.

 

[stewartLG]

I too wanted to express gratitude for this thread. I picked up a FS12-TY / C2100 from Deep Discount Servers. Cross flashing the H200 was deliberate but pain-free thanks to the guide you linked to, and after that it was just my usual tussle with getting FreeNAS to cooperate with Nagios. Deep Discount Servers charged me I'd guess a $75-$100 premium versus other vendors, but the box was clean, totally ready to go, and came with rails, so I have zero complaints.

The server is quiet enough for my purposes, haven't measured power draw. Better than my 4U 24 bay, that's for certain.

Thanks again.

 

[Holt Andrei Tiberiu]

I too wanted to express gratitude for this thread. I picked up a FS12-TY / C2100 from Deep Discount Servers. Cross flashing the H200 was deliberate but pain-free thanks to the guide you linked to, and after that it was just my usual tussle with getting FreeNAS to cooperate with Nagios. Deep Discount Servers charged me I'd guess a $75-$100 premium versus other vendors, but the box was clean, totally ready to go, and came with rails, so I have zero complaints.

The server is quiet enough for my purposes, haven't measured power draw. Better than my 4U 24 bay, that's for certain.

Thanks again.

Hello, i just bought a DELL R510 12 bay 2 u SERVER.
2 x XEON L5630, 128 GB DDR3 ECC, H200 - flash to IT mode in 20 min, 12 x 2TB Toshiba HDD's in, 2 PSU 80 GOLD, it has 2 internal USB POrts, like it is made for FreeNas

All of the serrver consumes ~120W to 130W, and it is cool and silent. It consumed 140W to 180W with the xeon E series cpu, but i changed them to L series.

 

[Redcoat]

I too wanted to express gratitude for this thread. I picked up a FS12-TY / C2100 from Deep Discount Servers. Cross flashing the H200 was deliberate but pain-free thanks to the guide you linked to, and after that it was just my usual tussle with getting FreeNAS to cooperate with Nagios. Deep Discount Servers charged me I'd guess a $75-$100 premium versus other vendors, but the box was clean, totally ready to go, and came with rails, so I have zero complaints.

The server is quiet enough for my purposes, haven't measured power draw. Better than my 4U 24 bay, that's for certain.

Thanks again.

I just got one too, from DFW as earlier commented in this thread. This is a completely new experience for me, having started with a FreeNAS Mini that came configured. So, I have the box on my desk, 6 4TB drives to put in it and 2 new 32GB Cruzer Fits. I downloaded the manual and will shortly take off the top and look around.
Thanks for the thread - I have a couple of questions lined up already so I'll be back once I get my bearings to ask you guys for some pointers...

 

[Redcoat]

OK, I'm feeling my way through this.

I have had the system powered up and set the boot to USB but have not installed FreeNAS yet. Using a USB hub I have a keyboard and two Cruzer sticks installed. I have another thumb drive with the FreeNAS image ready to go in a spare slot. I assume this is the right way to start out and end up with mirrored boot drives ...

Evaluating the system air flow I have concluded that I should not have any empty drive bays (drive caddies only) in use. So, given that I will burn in 8 drives for initial set up, then pull two of those out and have them as spares for this machine and my Mini, I propose to make 6 "dummy drives" of cardboard to sit in the otherwise open trays and distribute the air flow to create the same aerodynamic resistance as a full complement of 12 drives. Does this make sense to you experts or is this thinking flawed? Please comment.

The BMC web access freezes - say on selection of temperature sensors, for instance. Is this expected?

Oh, yes, do I run memtest86 with my drives in place, or before installing them and FreeNAS?

Thanks in advance for your advice.

 

[Spearfoot]

OK, I'm feeling my way through this.

I have had the system powered up and set the boot to USB but have not installed FreeNAS yet. Using a USB hub I have a keyboard and two Cruzer sticks installed. I have another thumb drive with the FreeNAS image ready to go in a spare slot. I assume this is the right way to start out and end up with mirrored boot drives ...

Evaluating the system air flow I have concluded that I should not have any empty drive bays (drive caddies only) in use. So, given that I will burn in 8 drives for initial set up, then pull two of those out and have them as spares for this machine and my Mini, I propose to make 6 "dummy drives" of cardboard to sit in the otherwise open trays and distribute the air flow to create the same aerodynamic resistance as a full complement of 12 drives. Does this make sense to you experts or is this thinking flawed? Please comment.

The BMC web access freezes - say on selection of temperature sensors, for instance. Is this expected?

Oh, yes, do I run memtest86 with my drives in place, or before installing them and FreeNAS?

Thanks in advance for your advice.

At the top of this thread, in post #4, @Mirfster gives detailed instructions on 'quieting the beast', which includes detailed recommendations about specific versions of BIOS and BMC firmware you should install on these machines.

The BMC web access freezing up is not good... Double-check your network connections and so forth.

I don't know how Dell's BMC works, but if it's similar to Supermicro's IPMI implementation, you should be able to mount the FreeNAS ISO locally on your workstation and pass it through to the Dell box in lieu of using a third USB drive connected directly to the box. But there's nothing wrong with using the later approach to install FreeNAS.

It won't do any harm to run memtest with your HDDs installed.

Good luck!

 

[Redcoat]

Thanks, Spearfoot. Memtest ran one pass so far just fine. Present frustration is to get the KVM working. Looks like I have a java security issue which I have not yet resolved - can't get the viewer file to run. So far not found anything helpful to solve it.
I'll put some drives in tonight and load FreeNAS.

 

[Redcoat]

Ahh - got KVM open in IE.

 

[BetYourBottom]

The BMC web access freezing up is not good... Double-check your network connections and so forth.

My BMC freezes as well when accessing certain things, like the sensors.

 

[Spearfoot]

@Redcoat , @BetYourBottom : did you flash the BIOS and BMC firmare to @Mirfster's recommended version? Curious to find out if doing so makes any difference in the stability and robustness of these systems. Thanks

 

[BetYourBottom]

@Redcoat , @BetYourBottom : did you flash the BIOS and BMC firmare to @Mirfster's recommended version? Curious to find out if doing so makes any difference in the stability and robustness of these systems. Thanks

I flashed the latest BIOS version and the recommended BMC version.

It's quiet as promised but opening the sensors window freezes the BMC and I have to refresh to be able to access other pages. However, that only happens if the machine is running, if you open the page while it's off it'll load all the sensor pages just fine and then after you start the machine you can refresh the pages to view the sensor data. It's hackneyed but it works at least a little bit.

 

[Redcoat]

@Redcoat , @BetYourBottom : did you flash the BIOS and BMC firmare to @Mirfster's recommended version? Curious to find out if doing so makes any difference in the stability and robustness of these systems. Thanks

Thanks very much for the response and followup. To your question, mine came from DFW configured as @Mirfster's recommended version (as far as I can determine). The behavior of BMC for me sounds to be much the same as @BetYourBottom's description above (I haven't tested all the combo's yet - certainly attempting to look at temps while the system is running causes the aberrant action, and there's at least one other menu selection that causes it). When I feel I understand it I'll post the information. I am interested to know if all our group of users involved here have the same experience and are ignoring it, or of there's something we should be attempting to address here...
I'm up and running with 8 4TB WD Reds installed and am about ready to start drive testing.
Once again, thanks for your continuing interest.

 

[Jason Hamilton]

I didn't have any issues with the temp sensors page on mine. I'll log in today and get the version number that I'm running and let you know what it is because my fans are quite. I honestly think that is quieter than my r720s that I have running for my esxi hosts.

Sent from my LG-H830 using Tapatalk

 

[Jason Hamilton]

Nope I lied mine does crap out on the sensors. Loads the fan page and then thats as far as it goes.

 

[Redcoat]

Nope I lied mine does crap out on the sensors. Loads the fan page and then thats as far as it goes.

Thanks for the confirmation. Let's see if anyone joins in with alternative experiences.