12-06-2013 12:01 AM
That is a shame. Guess I have to find somewhare outside the office for this piece.
Becuase I guess there is no solution for us running Debian with 4 disks configured as a software RAID5?
12-06-2013 01:21 AM
I have been following this thread as I also recently aquired a G8 MS and was shocked by the fan noise.
Regarding the resolution of this case, what a shame, I guess I will be returning my G8 MS and stick with the N40L a little longer.
I do hope HP will someday come to it's senses and develope a aciusticaly apropriate product as advertised for the small business / home use, as this product si clearly not in tune with such scenarios.
12-06-2013 08:45 AM
Yes I must admit this looks like a strange business decision coming from HP since the MicroServer line is marketed towards home and small office use.
Thanks to Jimmy Vance for helping us push this case and your support tickets directly through to the engineering department. I've been impressed with his level of care and dedication and for me this is what makes a big difference at the end of day.
For those of us utilizing the MicroServer Gen8 as a software-defined storage NAS, it seems the only solution would be to install an internal PCI-E SAS JBOD controller and connecting the 4-disks SAS backplane to it. This would allow installing a 2.5" HDD or SSD in the optical drive bay and configuring a RAID-0 array on the B120i to activate the iLO4 thermal sensor readout.
12-07-2013 02:12 PM - edited 12-07-2013 02:12 PM
@PAC_, @Jimmy Vance - does this mean we're at the end of the road for this problem?
Even if we were to accept that we have to use fake RAID (when are vendors going to realise that this is a dumb idea?), the HP driver is not Open Source and not available on the majority of Linux distributions. Surely there must be another workaround, such as a small bridging driver that pulls the data from the hard disks over AHCI and sends the thermal information to the iLO, or manually controlling the fan speed.
12-12-2013 02:38 PM - edited 12-12-2013 02:39 PM
After reading all of your posts in this thread I've become really disappointed at HP for the way they've treated this issue.
As HP obviously don't care about their customers I'm hoping to find a way around this. Has anyone tried adding a fan controller? Or changing to a more quiet fan?
12-13-2013 03:30 PM
Has anyone tried adding a fan controller?
-> yes, but as soon as the (ILO?) Controller recognizes a lower fan speed, it will boost the power and you end up having a loud server again
Or changing to a more quiet fan?
-> yes, but still its to noisy. On top of that, you have to split the cable as the connector is a non standardized one.
All in all, I am really upset because of this issue.
I've bought 2 (two) of these systems to have a small ESXi Testlab, running on 2 hosts.
The idea was to boot from USB, have an SSD in the optical as flash and 4 SATA disks in the bays.
But since no spin down of the drives occur, the servers are consuming loads power, eventually having an impact on the utility bill.
To be honest, after fiddling around with the first box and finding all of the above issues, finding this (and other) threads on this, I didn't even unpacked the second box.
So, if anyone is interested in buying two G8 Microservers, give me a shout.
p.s.: HP is also lacking more and more credibility in terms of support for "real" Servers. It's a disgrace how paying customers are treated these days. But thats (as always) just my thought on this. May be one clever Engineer (or BIOS hacker) will jump around the corner with an easy "fix" for this. (Recalling the issues on the G7 series' here).
12-13-2013 04:54 PM
@PAC_: Wish I could, but I said to myself, that I am not wasting any more time on this.
Hence, I did rip everything out and put the box back in the carton. Both are now advertised for sale elsewhere.
Its just to many minor issues adding up, eventually leaving me no choice then "giving up"..
Trust me, I'm not really happy with that myself, but its like within a project: When there is no chance of getting this done, pull the plug. So I did.
01-06-2014 02:35 AM
Question: Would 22% fan speed produce a noise level that is bearable in a home office environment?
I apologize for not making any updates in this thread for some time, but I don't have any official update information to provide.
As far as this question goes, I don't know that there is an easy answer. As you can see in this thread some have said it is too loud, in my home office (which isn't really all that quiet) it is fine. I did drop off a system at a friends house to get another opinion. His ambient noise level is very low. With the unit sitting on the floor next to his desktop, neither of us noticed any fan noise. He really likes the MicroServer and as such, I'm having a hard time getting it returned :)
No support by private messages. Please ask the forum! I work for HP
If you feel this was helpful please click the KUDOS! thumb below!
01-06-2014 04:17 AM - edited 01-07-2014 07:23 AM
"Question: Would 22% fan speed produce a noise level that is bearable in a home office environment"
not ideal but that should still easily provide enough airflow whilst reducing the fan noise to a reasonably acceptable level (at 6% the server is very quiet, at 36% it's so loud I can easily hear it at the other side of the house - I'd hope 22% would be in the 'loud but reasonable for a small server' range). If that was the best HP could offer us I could compromise at that figure I guess but I still think it would be quite easy to improve on that, much better than nothing though which is what we have now.
01-06-2014 05:55 AM
"I apologize for not making any updates in this thread for some time, but I don't have any official update information to provide.
As far as this question goes, I don't know that there is an easy answer. As you can see in this thread some have said it is too loud, in my home office (which isn't really all that quiet) it is fine. I did drop off a system at a friends house to get another opinion. His ambient noise level is very low. With the unit sitting on the floor next to his desktop, neither of us noticed any fan noise. He really likes the MicroServer and as such, I'm having a hard time getting it returned :)"
Seriously? I can only assume that you live in a very noisy area or that your hearing isn't what it once was (or you're using the raid modes which are known to be quiet), in AHCI mode the server is extremely loud and can be heard from many meters away in different rooms of my house even with doors closed, in the same room it's totally unacceptable - you can't hear yourself think with this thing going! If I had two of these servers running it would be about as loud as a standard vacuum cleaner, thankfully I only bought one!
It's taken me quite a while to calm down enough to post on here since HP posted their "can't be bothered to fix this" response, it's totally unacceptable in it's current form and fixing it really shouldn't be a big problem. I can understand that without HDD sensor data they want to play a bit safe with fan speeds but the 'safe' speeds they've set are ridiculously high.
I'd really like someone from HP to answer this question:
Since you have a 'sea of sensors' and temperature data from all areas around the hard drive cage why can this data not be used to work out when to increase the fan speed?
If you put the drives in raid mode then gave the drives lots of work to do you could easily see what sort of temperature change happens around the drive cage based on the drive temperature changes, working out a more sensible 'safe' speed to run the fans at would be easy from this info.
Eg something like:
temp of sensors around drive cage = same as startup temp = disks obviously cool = fans at 6%
temp of sensors around drive cage = startup temp +5 degrees = disks warming slightly = fans at 15%
temp of sensors around drive cage = startup temp +10 degrees = disks getting warm = fans at 25%
temp of sensors around drive cage = startup temp +15 degrees = disks getting hot = fans at 40%
Obviously adjust degree changes/fan speeds to sensible levels based on above testing but this should be barely an afternoon of testing/adjusting to come up with a way of keeping the drives cool whilst not just defaulting to crazily high fan speeds at all times for no reason. By all means use the hottest disks you can find for testing and add a little safety margin to the numbers but for HP to just ignore the issue and hide behind a 'fans have to run at near 40% at all times for safety' response is pathetic, a multi billion dollar company with all the resources they have, incredibly bright staff and full knowledge of the iLo's capabilities should be able to do much, much better - it just proves they've barely spent a couple of minutes considering the problem before taking the easy way out which is what really irritates me the most. I feel this problem unfortunately has a lot more to do with HP being lazy than actually being unfixable, or maybe someone from HP can properly explain otherwise?
01-07-2014 09:43 AM
50 dBa is very loud for a home office environment. I have to say the noise bothers me quite a lot as well since it is the only equipment producing background noise in our studio.
The Apple iMacs were inaudible at about 21 dBa. Just for kicks, the recently announced Apple Mac Pro produces 12dBa (!!) at idle.
I have done some additional testing in regards to the SmartArray B120i. Even with a RAID-0 array configured, iLO 4 loses HDD sensor data once FreeBSD is booted.
I have yet to test with ESXi, but this means the SmartArray would need to be active with the driver present and loaded within the OS in order for the fans to speed down.
01-07-2014 09:11 PM - edited 01-07-2014 09:13 PM
A few comments after reading this thread and following along.
My background, 25+ year computer professional, EE, have been a unix/network admin in some capacity or another for 20 years. Worked for HP back in the late 90's/early 2000's on CPU design. Now a small business owner. I maintain an IT infrastructure for a few small companies that currently involes maybe a dozen servers and workstations, and a bunch of macs and laptops for office use. The workstations we have are z620's and z420's. Some are used as workstations, some as servers or mixed workloads. We have an ML110/G7 and one Microserver. Most of our network hardware is HP.
I recently bought a MicroServer Gen8 for our office, to run a business-operations database (linux). It runs off an SSD in bay 1 with additional software-defined redundancy on two disks in bays 2 and 3. (IE, it's not configured as a "storage server"-- we have z620's and the ML110 set up for that.)
Overall a very nice package. Once booted into Debian 7.3, the fans run at about 32% when there is no or minimal load. Inlent ambient at 21C, 12-Sys exhaust at 37C. CPU @ 40C. The fans are audible in a reasonably quiet office environment. I would not choose to put this "under my desk." In contrast, the z620 I have sitting 18" behind my monitors is extremely quiet- like I have to put my ear within 12" to hear it at all. I would swear that it runs fanless most of the time, but maybe they're just super quiet.
And yes, when you get the z620 humming with a lot of work to do, it gets louder. I don't think anybody expects no fan noise when there is a lot of work being done. But most of these machines have a very low workload compared to their peak/burst performance 99.9% of the time.
I'm not too wrapped around the axle about the fan sound from the Microserver, but it would make me want to place the server somewhere where I didn't care about noise, like the router closet. Definitely not something I want to put in my personal workspace or under someone's desk in a quiet office or room. I would not want it running in a home theater, etc.
The MicroServer Gen8 has clearly been designed with cosmetics in mind, and I admit it does look really cool. HP has clearly put some effort into making the Gen8 series "look great", and they have obviously put a lot of effort into making the MicroServer Gen8 as compact as possible-- which is appreciated. It's a really nice package.
But this brings up a disconnect in my mind. If the MS is not going to be quiet enough to put on the desk, under the desk, on the A/V rack with the stereo, or on the bookshelf, then it'll be relegated to some place out of the way like the server closet. And then who cares if it looks awesome.
And more importantly, who cares that it's smaller than a ML110, or, what appears to be its successor, the ML310e. For about the same real price (I use Provantage for comparisons), you can get an ML310e. And that comes with an E3-1220V2, 4GB, the same 4x 3.5" bays, but also two external 5.25" bays and 4 PCI slots. If you want to run a bunch of external storage via an SAS/SATA card over SFF-8087-- or even a real internal RAID card, and then, well, anything else, you have the slots for it. For the guys into virtualization you could dedicate extra NICs to VMs or whatever.
The prior poster (ad-d) is right, if we've gt the "Sea of Sensors", HP should be able to use that data to infer drive temperatures and cooling. Heck, the z620 I have on my desk doesn't have any of the funky RAID or "Sea of Sensors" stuff, and I've got 6 hard drives running at all times in it. Works great and is very quiet.
Paulgear is dead on that fake raid is a joke. If I am going to run software RAID, it's going to be ZFS on Linux. I am not going to depend on some weird hardware/software raid driver such that I cannot take the drives out of the system, plug them into any other system with 4 SATA ports, and attempt to bring up the array or do data recovery.
Furthermore, the idea that we should put individual drives into "fake" 1-drive RAID0 arrays to fool the controller into thinking they're in a RAID just so the cooling stuff works right is ludicrous. For the same reason. I don't want to be required to use a special controller to get data off of it if I have to move it to a different machine, nor do I necessarily want to have to initialize all the disks I individually add to the system with a fake raid marker on the drive.
The MicroServer wil never be right for big compute jobs given the thermal design and max TDP. But what it excells at is being small, having good connectivity (2xGigE) and ILO, and having a lot of storage slots available in a very small physical package. Heck, you'd be hard pressed to buy a dedicated NAS RAID box from a major vendor (Synology etc) with 4x 3.5" bays for the same cost. So run FreeNAS or Linux on it and you're ahead.
So all that said, here are some concrete suggestions for HP, in no particular order:
(1) Give us all the SATA ports that the Chipset supports. It would be awesome for these to all be 6 GBbps (or faster) but realistically the chipsets rarely support the fast speed on all ports. Make them all equal and bootable, don't require us to use a legacy mode to boot from 1 or another of the ports.
(2) CD/DVD's are pretty much obsolete for computer media. They are still used to disrtibuted commercial software for people who do not have a fast internet connection, and for people who want to watch DVDs on their computer. So use of one of the extra SATA ports for a DVD drive should be supported, but not intended as its primary purpose. The natural thing to do in the MicroServer Gen8 is to throw 1 or 2, 2.5" SSD's, on the top of the cage when you take the cover off, and to boot from one of those. So make that a supported configuration. It just involves not adding artificial limits to the boot order and AHCI vs. Legacy-IDE mode, a few sheet metal changes, and more standard cabling up there.
(3) Don't make us waste a front-pluggable drive bay on a boot/root/OS drive. That's just dumb. Jeez, almost everyone on here wants to boot from an SSD, run the OS from it, and put an array of 4x 2-, 3-, or 4 TB drives in a RAID5, RAIDZ, or RAID10 over the 4 drives in the bays.
(4) "Fake raid" (that is, a raid controller that requires OS support to actually do RAID) is lame. It's OK if you want to offer it, but don't assume that it's the preferred solution and put barriers in that affect us using proper OS/filesystem level redundancy, ie, ZFS.
(5) Fix the fans so the system is quiet. I know you want to know the drive temperature of everything but you have a "Sea of Sensors" to figure that out, and requiring use of your fake raid controller instead of a proper software RAID/redundancy solution to make the fans work is dumb.
(6) Build a 4, 6, or 8-bay storage expansion box that is styled after the MicroServer Gen8 and stacks on top of it. The MicroServer was built to stack. As something very small with at least 4 drive bays, it was also clearly designed to do storage. There are plenty of SATA/SAS controllers (HP, LSI, or otherwise) that support 4 or 8 external SATA drives or more than that SAS drives. It would be great to buy a MicroServer, throw a 4x or 6x 3.5" bay expansion box that stacks right on top of it, and then run the whole 8, 10, 12, etc drive array on ZFS.
01-08-2014 12:20 AM
Whilst my exact needs and preferences differ slightly from yours, the sentiment is the same: HP could have made this server the go-to device for nearly every small business on the planet. But because of the poor decision-making process surrounding the fakeRAID driver and the noise, it won't be.
I'm working with my main client at the moment on a re-vamp of their remote office architecture, and we have just discarded the MicroServer Gen8 as an option because of the fakeRAID/AHCI noise issue. Otherwise it would have been perfect. But most of our remote offices don't have a dedicated server room, and for the health and safety of our staff we can't put these servers next to them. (On that note, HP Networking really needs to release a viable competitor to the Juniper EX2200-C for remote offices: 12-port Gigabit, 2-port SFP, fanless, Comware OS, PoE as an option.)
01-08-2014 12:28 AM
01-08-2014 07:06 AM
After I read this thread over the weekend, I decided to test the Microserver Gen8 (codename LEO) in my lab. I've been debugging the iLO4 health subsytstem while the server runs in SATA AHCI mode and I know exactly what is going on here and I also know what can be done to solve this fan noise issue.
I've talked with people in the iLO team but, we are going to need input from the thermal guys before we can move forward.
I also talked to Jimmy Vance earlier this week and he told me he's still in contact with the LEO team and actively looking for a solution from his side as well.
01-08-2014 09:36 AM
I forgot one more thing
(7) Make the darn thing boot faster. I don't mean the OS part. Linux boots off an SSD in just a few seconds. I mean getting through all the BIOS and into the boot loader. It takes forever on the Proliants to boot. I know it's a server and I leave all my machines on and have long uptimes. However, if we need to reboot something "real quick" for whatever reason, it's a pain to have it down for what seems like 5 minutes. It's probably shorter than that but it takes too long.
01-08-2014 01:09 PM
I was under the impression after PAC_'s post (http://h30499.www3.hp.com/t5/ProLiant-Servers-Nets