Dell H730 mini ... I'm confused

[ChipP]

Hi Folks.

Looking back through discussions re- Dell H730, it seems that way back, consensus was that putting the H730 into HBA mode "wasn't really" the same as having a JBOD HBA and "bad things" would happen if you did.

More recently I see some folks are having good results.

I've tried h730 [and h740] in a test rig, and they seem to work okay.

Really simple question ... is running FreeNAS on a Dell H730 in HBA mode still a stupid thing to do?

Thanks for your patience

Chip.

 

[Chris Moore]

a Dell H730 in HBA mode

I just don't understand why you would want to do that.

Take a look at some of the videos this guy makes. He has had some interesting results with flashing controllers to work in Dell servers:

Dell H310 mini card with LSI IT mode firmware
https://www.youtube.com/watch?v=Y1Xi5NZRlXM&feature=youtu.be

How to fix: "Invalid PCIe card found in the Internal Storage slot!" error in Dell PowerEdge R710
https://www.youtube.com/watch?v=v0AEHVdc_go

I can't see YouTube from work, but I know he has done a lot of more recent work with the Dell platform hardware.

 

[Chris Moore]

still a stupid thing to do?

It isn't so much stupid as potentially harmful to your data. The problem that was happening is that the hardware RAID card, even in HBA mode, was preventing FreeNAS from being able to see accurate status of the disks. This prevented FreeNAS from alerting you to hardware faults that might impact your data. It will work, until it doesn't and the first notification that it isn't working any more might be a catastrophic data loss.

 

[ChipP]

I just don't understand why you would want to do that.

Take a look at some of the videos this guy makes. He has had some interesting results with flashing controllers to work in Dell servers:

Dell H310 mini card with LSI IT mode firmware
https://www.youtube.com/watch?v=Y1Xi5NZRlXM&feature=youtu.be

How to fix: "Invalid PCIe card found in the Internal Storage slot!" error in Dell PowerEdge R710
https://www.youtube.com/watch?v=v0AEHVdc_go

I can't see YouTube from work, but I know he has done a lot of more recent work with the Dell platform hardware.

Thanks for your time Chris.

Why? ... simplez ... the r730 I have came with a H730 mini it it and this seemed like a reasonable place to start.

Thanks for the reference to the videos. They are great, aren't they? I have a 620 with a 310 in it I flashed myself and that's worked faultlessly.

I haven't been able to find the equivalent for flashing a H730 though, in fact a couple of people have said this isn't actually possible. I don't suppose you happen to know if tis is possible?

Once again - thanks for taking the time on this. Really appreciate it.

CHip.

 

[ChipP]

It isn't so much stupid as potentially harmful to your data. The problem that was happening is that the hardware RAID card, even in HBA mode, was preventing FreeNAS from being able to see accurate status of the disks. This prevented FreeNAS from alerting you to hardware faults that might impact your data. It will work, until it doesn't and the first notification that it isn't working any more might be a catastrophic data loss.

Ah - okay. Based upon the fact that I could access the SMART info and all that malarkey I'd assumed that FreeNAS had everything it needed to do its thing.

If that isn't the case then what you and many others are saying makes complete sense.

Thanks again

CHip.

 

[rvassar]

Just for info backfill... The more modern Dell PERC cards implement something called "ePD-PT" rather than a JBOD, and they reuse the terminology in Perccli and the host BIOS configuration interfaces. These ePD-PT's can make use of some controller features under a variety of circumstances, while maintaining the illusion of a disk directly attached to the host.

 

[ChipP]

Thanks.

I'm beginning to think line of least resistance might be to pick up a HBA330 [or H330 and flash my own].

 

[Chris Moore]

Thanks.

I'm beginning to think line of least resistance might be to pick up a HBA330 [or H330 and flash my own].

The guy that makes those videos also sells on eBay. Just search for, "SAS HBA IT". He always has the wooden tabletop behind his cards. I have bought from him and the cards were good, working as described and not unreasonably priced.

 

[ChipP]

Hi Folks,

Thanks for all the help with this.

JiC anyone is wondering ...

I managed to get my hands on a H330 from an e-bayer about 15 mins drive from me. Re-flashed that and its running fine [or at least ... it seems to be].

If I have any problems moving forward I'll be sure to let you all know.

Thanks again

CHip

 

[aeubank]

Just reviving an older thread (which from what I believe is preferred) I have a T630 server with a H730 RAID card in it set to HBA mode.

I purchased an HBA330 card expecting to have better uses for TrueNas, BUT there was one major side effect. The fan speed increase..
The fan speed noise increase is very noticeable with the HBA330 card. I even found a support article that talks about this.
So, I'm going ADO (against doctor's orders) and keeping the H730 RAID card in HBA mode in the computer as this computer sits in a room and additional noise would make life difficult.

My plans are to check the SMART settings and everything else closely to make sure there is no data loss.

Does anyone see additional side effects to this? This configuration was pretty much standard back in the day, so is there anyone else running this same configuration successfully?

Thanks
aeubank

 

[rvassar]

Unplug the battery.

 

[rafalefighter]

Unplug the battery.

Will this work on HPE p440ar as well ? If I unplug the battery and use the card in HBA mode ?


Also, based on this thread I really don't get what is the bad part using raid card in HBA mode. is that only due to missing smart data to the Truenas ?

The card I'm testing right now (p440ar) in HBA mode present all the data and raw disk to the Truenas. I can see SMART data from Truenas. Plus we can monitor all the drive status via SNMP through HPE ILO which alert if the drive failing right away.

If SMART data is the only missing part not sure why need to use different HBA. But if the driver compatibility is the issue then that's differnt story

 

[rvassar]

Will this work on HPE p440ar as well ? If I unplug the battery and use the card in HBA mode ?


Also, based on this thread I really don't get what is the bad part using raid card in HBA mode. is that only due to missing smart data to the Truenas ?

The card I'm testing right now (p440ar) in HBA mode present all the data and raw disk to the Truenas. I can see SMART data from Truenas. Plus we can monitor all the drive status via SNMP through HPE ILO which alert if the drive failing right away.

If SMART data is the only missing part not sure why need to use different HBA. But if the driver compatibility is the issue then that's differnt story

I'm not familiar with the HPE p440ar, so I can't give you an accurate answer.

The problem with the newer RAID cards is they've started overloading their "jbod" tagged disks. Several controllers end up creating hybrid devices where the OS thinks it has a direct attach disk, but the controller is still using it's RAM cache to optimize the I/O according to what the controller FW thinks is best. This creates problems:

1. The performance results are misleading, and ZFS will make incorrect assumptions that can have gross performance effects.
2. An undetected "pinned cache" event can trash your pool. If you hit an error and move the drive to an unused non-controller port, you've might import a drive that matches the ZFS UUID for the pool, but is missing whatever data is stuck on the controller cache. ZFS may detect this these days, but I suspect not.

Unplugging the battery on some controllers makes a cache offload event is unsustainable, and forces all drives to write-thru. In theory...

 

[rafalefighter]

Thanks for the details.

Is there any way to verify this after I unplug the battery ? Or may be verify that controller presenting real drive to the Truenas but do optimization in back end on the fly ?

And is this valid even if all my drives presented with correct serial numbers and other data like SMART and temp readings ?

 

[rvassar]

Thanks for the details.

Is there any way to verify this after I unplug the battery ? Or may be verify that controller presenting real drive to the Truenas but do optimization in back end on the fly ?

And is this valid even if all my drives presented with correct serial numbers and other data like SMART and temp readings ?

Again, I'm not familiar with your card. Verifying this for an end user... It might generate a system event warning all writes are "write thru" or something like that. It might even generate a boot event that has has to be manually acknowledged. It kind of depends on the server and the options configured.

 

[rafalefighter]

I'm sorry, what I mean is there a way to verify this on Truenas End ? that there's no middle layer between freebsd and the disk ?

For the server end, we do have few servers with p440ar smart storage battery degraded. Server keep telling it while post and also from snmp. And IPMI within the Truenas as well.

I would like to test Truenas on one of these servers with dead battery. To make sure without the battery p440ar not doing these optimizations.