|
IT Guy posted:gently caress, sorry for all the questions. Are they advanced format drives with physical 4K sectors? Then yes, go for it. If not, things could get a bit screwy messing with that, so leave it.
|
#
?
Mar 26, 2012 20:57
|
|
|
|
| # ? Apr 25, 2024 16:17 |
|
|
thideras posted:That would certainly require a lot of force and would kill the board instantly! Most likely not. If I remember correctly, 6+2 and EPS have the same physical connector. They DO however have GND and +12V flipped. As long as you have the appropriate ATX pin remover, that's not hard to fix though.
|
|
#
?
Mar 26, 2012 21:00
|
|
|
IT Guy posted:I wanted to use that slot for a 60GB SSD ZFS cache I had laying around.  ) something on ZFS, but it largely depends on the disks you have and their usage. Why not look at the performance without doing any tweaking first? I'm maxing out my LAGG'd NICs over SMB no problem with read and write is around 170MBps and that's without cache or zil drives and on 4x Samsung HD204UI 4k sector drives. Honestly, ZFS is good out of the box.Also, if you want both cache and zil, you want at least 3 ssd disks (1 at the same size as your total physical memory for cache, 2 at ~20GB each for zil in a (z)raid1 to avoid potential dataloss). As to your question regarding 4k sectors vs 512b sectors, it only matters if drives you're purchasing in the future which won't have 512b emulation (which I don't ever see happening, ever, personally). 4k sector drives are said to out-perform 512b sector drives, but it heavily depends on the load in question. If you don't want the worry about it (as you can't change mix 512b with 4k sectors in one vdev, just leave it at 512b. If you're ever adding more vdevs, you can have them run 4k sectors just fine. Incidentally, have you read the ZFS Best Practices Guide or the ZFS Evil Tuning Guide? You really should if you're looking to do performance optimization without doing serious testing first and making sure you don't have any bottlenecks / wrong default-configuration. Sorry, the post became a bit cluttered in adding on information as I thought of it. BlankSystemDaemon fucked around with this message at 21:34 on Mar 26, 2012 |
|
#
?
Mar 26, 2012 21:23
|
|
|
D. Ebdrup posted:I'll definitely give that a read. Thanks.
|
|
#
?
Mar 26, 2012 22:00
|
|
|
D. Ebdrup posted:2 at ~20GB each for zil in a (z)raid1 to avoid potential dataloss).
|
|
#
?
Mar 27, 2012 03:32
|
|
|
FISHMANPET posted:For what it's worth, that's not a "barebones" server in the traditional meaning of the word barebones when it comes to computer, that's a NAS appliance without any disks. As to your question, ff you're looking at an appliance, then it depends on the appliance if you can expand or not. If you're talking about true barebones (like the HP Microserver) then it depends on what OS you run. Ah, OK. Got the terminology mixed up there. I guess I'm looking for an appliance then if a NAS is a recommended choice. Would a NAS be the cheapest and most reliable solution to store ~20TB of data? After doing a little research on Google, people recommend tape drives and tapes for storage too as well as using online storage sites like Carbonite to save a second set of that data should the hardware fail. $60/year to store an unlimited amount of data seems like a bargain.
|
|
#
?
Mar 27, 2012 03:32
|
|
|
Charles Martel posted:Ah, OK. Got the terminology mixed up there. I guess I'm looking for an appliance then if a NAS is a recommended choice. Is there even a single NAS that can store 20tb? If you want that much data just get a norco server box.
|
|
#
?
Mar 27, 2012 04:07
|
|
|
kri kri posted:Is there even a single NAS that can store 20tb? edit: Any 8-drive 3 TB RAID 5 array should be 21 TB, though at that point I'd be tempted to use RAID 6.
|
|
#
?
Mar 27, 2012 04:10
|
|
|
QNAP makes an 10 bay unit (also 12 bay, but rack mount only). http://www.qnap.com/pro_detail_feature.asp?p_id=200 We just got one of those at work to do some testing with.
|
|
#
?
Mar 27, 2012 04:19
|
|
|
Charles Martel posted:Would a NAS be the cheapest and most reliable solution to store ~20TB of data? After doing a little research on Google, people recommend tape drives and tapes for storage too as well as using online storage sites like Carbonite to save a second set of that data should the hardware fail. $60/year to store an unlimited amount of data seems like a bargain. For online storage, you may want to add Backblaze and, my personal favorite, CrashPlan. Backblaze is ~$4/mo if you buy a year and CrashPlan is $3/mo. CrashPlan works on Linux (which is why I use it) and you can have other computers or your friends backup to you for free. Good luck uploading 20tb of data and keeping it current. I exclude all media that can be reproduced (ISO's, video/audio), which leaves me around 10gb of data to keep updated. EDIT: Regarding that TS-1079 Pro, I'd rather just build my own for cheaper and be able to do whatever I want on it (VM, etc). That thing is expensive. thideras fucked around with this message at 04:23 on Mar 27, 2012 |
|
#
?
Mar 27, 2012 04:20
|
|
|
I setup 4x 2TB drives in a zpool with RAIDz1. Before creating any datasets or writing any data to the array, my size was 5.3TB total which is correct. After creating 3 Datasets and rsyncing data to them, my sizes look like this:  I'm a first time user of ZFS, am I doing this correctly? Should I be using datasets or should I just use the main mount point? The only reason I used datasets was so I could share the 3 different folders. Why are my sizes all hosed? Edit: After some investigating, I've found that this behaviour is a known issue and is an open ticket. IT Guy fucked around with this message at 16:49 on Mar 27, 2012 |
|
#
?
Mar 27, 2012 16:12
|
|
|
I'm gonna put the 250gb drive that comes with my N40L into the 5.25" bay, but I'd like to secure it. Do you guys know of any 3.5"-in-5.25" devices?
|
|
#
?
Mar 27, 2012 18:15
|
|
|
Any hardware reseller should have brackets for less than $10
|
|
#
?
Mar 27, 2012 18:25
|
|
|
thideras posted:With that much space, you are either going to be making your own server or buying [really expensive] 4tb disks. Tape drives would be good for backups or files that you don't need to access "right now" and that rarely/never change. I'd rather have the spinning disks, though. Does nothing back up NAS drive? From what I can tell, setting up an iSCSI would not work, since I want to be able to copy files onto my NAS back up from different computers. Are there any solutions out there that I can just drag and drop files manually to be stored or is it all through some program that you have to set what to back up? I'm running FreeNAS.
|
|
#
?
Mar 27, 2012 18:27
|
|
|
Fangs404 posted:I'm gonna put the 250gb drive that comes with my N40L into the 5.25" bay, but I'd like to secure it. Do you guys know of any 3.5"-in-5.25" devices? http://www.amazon.com/StarTech-com-Universal-Mounting-Bracket-BRACKET/dp/B0001UZQWG Or google "5.25 to 3.5 adapter". The ones I lined are cheap as poo poo but if you forsee needing to change the drive out it's a bit of a PITA.
|
|
#
?
Mar 27, 2012 18:29
|
|
|
Fangs404 posted:I'm gonna put the 250gb drive that comes with my N40L into the 5.25" bay, but I'd like to secure it. Do you guys know of any 3.5"-in-5.25" devices? http://www.officedepot.com/a/products/141995/StarTechcom-35in-Universal-Hard-Drive-Mounting/
|
|
#
?
Mar 27, 2012 18:29
|
|
|
Prefect Six posted:Does nothing back up NAS drive? As long as your FreeNAS server is the host to the iSCSI volume, why wouldn't that work? You'd treat the iSCSI array as another storage unit, though the NAS.
|
|
#
?
Mar 27, 2012 18:35
|
|
|
Prefect Six posted:Does nothing back up NAS drive?
|
|
#
?
Mar 27, 2012 18:49
|
|
|
Civil posted:As long as your FreeNAS server is the host to the iSCSI volume, why wouldn't that work? You'd treat the iSCSI array as another storage unit, though the NAS. From what I could tell only one PC could access the iSCSI volume, so my wife couldn't upload pictures from her laptop onto the iSCSI volume through a SMB share if the iSCSI initiator was on my PC. This is just what I've gleened from looking into setting up an iSCSI connection, I could be completely wrong so please educate me.
|
|
#
?
Mar 27, 2012 18:52
|
|
|
thideras posted:I don't quite understand what you are asking. You can access a NAS server with NFS, SSH, Samba, iSCSI, etc. If you are using Windows, you can simply create a Samba share and map the drive on the client. There should be no software required. All of the back up clients I have seen say no network shares on the website. Granted I haven't actually done a trial to see if the client would see a network device. I have the NAS set up on an SMB share currently.
|
|
#
?
Mar 27, 2012 18:54
|
|
|
Prefect Six posted:From what I could tell only one PC could access the iSCSI volume, so my wife couldn't upload pictures from her laptop onto the iSCSI volume through a SMB share if the iSCSI initiator was on my PC. Think of iSCSI as a USB connection to a drive. Connect it through the freenas machine, then share it from there.
|
|
#
?
Mar 27, 2012 18:54
|
|
|
ISCSI is a block protocol (so if you're going to mount it on n>1 machines you'll need a concurrent access filesystem, which NTFS isn't. Share your NAS volumes using a file protocol (smb/cifs, nfs, afs, you name it). Then the server process takes care of locking etc
|
|
#
?
Mar 27, 2012 18:56
|
|
|
Civil and evil_bunnY beat me to it. Just share it out with Samba.
|
|
#
?
Mar 27, 2012 18:58
|
|
|
So once I have the iSCSI volume set up I can share it with the SMB protocol? This is probably what I missed and why I'm so confused.
|
|
#
?
Mar 27, 2012 18:58
|
|
|
Prefect Six posted:So once I have the iSCSI volume set up I can share it with the SMB protocol? iSCSI is for mounting block-level devices over a network. You do not need that.
|
|
#
?
Mar 27, 2012 19:00
|
|
|
thideras posted:Where is iSCSI coming into all this? Mount the drives and share with SMB. I guess I was under the impression that using iSCSI was a way to get around the 'no-NAS' rule of most popular back-up clients. If I can map the drive through windows and have the back up clients detect that then, yeah that would be a much cleaner, simpler solution.
|
|
#
?
Mar 27, 2012 19:09
|
|
|
Ceros_X posted:http://www.amazon.com/StarTech-com-Universal-Mounting-Bracket-BRACKET/dp/B0001UZQWG Civil posted:They're everywhere. Same concept as the 3.5" to 2.5" adapters. Thanks guys!
|
|
#
?
Mar 27, 2012 19:14
|
|
|
Prefect Six posted:I guess I was under the impression that using iSCSI was a way to get around the 'no-NAS' rule of most popular back-up clients. If I can map the drive through windows and have the back up clients detect that then, yeah that would be a much cleaner, simpler solution.
|
|
#
?
Mar 27, 2012 19:15
|
|
|
thideras posted:It will depend on what software you use. If this is just local backups, no software should complain about backing up over the network -- that would be pretty stupid. Backing up online, however, I could see where they don't want that. I only have Linux experience with CrashPlan, but I'm pretty sure that you can backup whatever you want. When I ran their software on my main desktop, it allowed me to backup my network share. Whether or not they check for it in Windows, I don't know. Right, I'm trying to back up what's on my NAS to the Internet as an off-site back up. I'm not familiar at all with Linux and not sure if their Linux client would run on FreeBSD or not. Regardless, I'll check out crash plan and see what happens.
|
|
#
?
Mar 27, 2012 19:18
|
|
|
Quote != edit
|
|
#
?
Mar 27, 2012 19:19
|
|
|
So I'm building one of the HP microservers, and I ordered 5 2TB samsung drives that I was planning on running in RaidZ1. Then I got all paranoid because I read some articles that seem to imply that if you lose a drive you WILL end up with a completely corrupted array when you try to rebuild the array due to unrecoverable read errors. I bought another drive, for RaidZ2. This server is going to be used for backups and storing media. I'm asking some of you guys with experience, do I need RaidZ2, or should I just use RaidZ1 and get 2tb more space out of the deal?
|
|
#
?
Mar 27, 2012 20:20
|
|
|
UndyingShadow posted:So I'm building one of the HP microservers, and I ordered 5 2TB samsung drives that I was planning on running in RaidZ1. Then I got all paranoid because I read some articles that seem to imply that if you lose a drive you WILL end up with a completely corrupted array when you try to rebuild the array due to unrecoverable read errors. I bought another drive, for RaidZ2. RAIDz1 allows for one drive failure, can you link the articles that said otherwise?
|
|
#
?
Mar 27, 2012 20:25
|
|
|
IT Guy posted:RAIDz1 allows for one drive failure, can you link the articles that said otherwise? http://www.zdnet.com/blog/storage/why-raid-5-stops-working-in-2009/162
|
|
#
?
Mar 27, 2012 20:38
|
|
|
UndyingShadow posted:http://www.zdnet.com/blog/storage/why-raid-5-stops-working-in-2009/162 Interesting. I don't know what to tell you. I guess I'm going to go turn my backup schedule on more frequently now.
|
|
#
?
Mar 27, 2012 20:57
|
|
|
UndyingShadow posted:So I'm building one of the HP microservers, and I ordered 5 2TB samsung drives that I was planning on running in RaidZ1. Then I got all paranoid because I read some articles that seem to imply that if you lose a drive you WILL end up with a completely corrupted array when you try to rebuild the array due to unrecoverable read errors. I bought another drive, for RaidZ2.
|
|
#
?
Mar 27, 2012 21:13
|
|
|
UndyingShadow posted:So I'm building one of the HP microservers, and I ordered 5 2TB samsung drives that I was planning on running in RaidZ1. Then I got all paranoid because I read some articles that seem to imply that if you lose a drive you WILL end up with a completely corrupted array when you try to rebuild the array due to unrecoverable read errors. I bought another drive, for RaidZ2. Just checking but by my count that is 6 drives you have ordered for a 4-bay server. Are you aware of this and doing some mod/chassis-hacks?
|
|
#
?
Mar 27, 2012 21:52
|
|
|
UndyingShadow posted:http://www.zdnet.com/blog/storage/why-raid-5-stops-working-in-2009/162 One of the nice features of ZFS is that it's aware of what's on the drive, so when it reslivers (the term for a ZFS rebuild) onto the new drive, it only reslivers actual data, not the entire drives worth of nonsense. So that problem doesn't really come into play until you have 12TB of actual data (assuming the URE rate is still 10^14). E: I just checked the Hitachi 7k3000, the consumer drives are still 10^14, and the Enterprise are 10^15, just like he says. But with ZFS it also matters what a URE means. If it's an actual bad sector then you've just lost whatever file was in that sector, but you may be able to clear the error and read the sector happily again. FISHMANPET fucked around with this message at 22:42 on Mar 27, 2012 |
|
#
?
Mar 27, 2012 22:37
|
|
|
zacd posted:Just checking but by my count that is 6 drives you have ordered for a 4-bay server. Are you aware of this and doing some mod/chassis-hacks? The internal sata port for the CD drive can be hacked with a bios to turn it full speed, then an esata to internal sata cable can be run from the back of the machine back in to the top. Then with a special bracket, you can put 2 hard drives in the 5 1/4 bay for a total of 6.
|
|
#
?
Mar 27, 2012 23:07
|
|
|
FISHMANPET posted:One of the nice features of ZFS is that it's aware of what's on the drive, so when it reslivers (the term for a ZFS rebuild) onto the new drive, it only reslivers actual data, not the entire drives worth of nonsense. So that problem doesn't really come into play until you have 12TB of actual data (assuming the URE rate is still 10^14). That makes me feel better. The article made it seem that you would lose the entire array (maybe that's the case for raid 6, since it doesn't actually seem to know anything about the data on the drive) I could handle a lost file, but the thought of losing all the data due to a URE was a little hard to deal with. Thanks for the info.
|
|
#
?
Mar 27, 2012 23:10
|
|
|
|
| # ? Apr 25, 2024 16:17 |
|
|
IT Guy posted:Edit: After some investigating, I've found that this behaviour is a known issue and is an open ticket. As for a single disk failure causing data loss, I'll paraphrase the zdnet guy. Disks have a specified unrecoverable read error rate. Consumer SATA drives have so much capacity, that it is likely you will encounter an unrecoverable read error at one point during a rebuild of a sufficiently large array. Without the parity data during a rebuild, you have the potential for data loss. Most disks will go into "Heroic Recovery Mode" which exasperates the situation, because the disk will appear unresponsive for seconds to minutes, which could cause it to drop out of the array. So not only do you have to contend with not being able to trust your data after a full rebuild, you also have controllers that will flat out stop a rebuild on a URE with no additional parity data, and you also have to deal with the appearance of a 2 disk failure (though the other disk will eventually come back) knocking your array offline. I use raid 5 for 5 or less disks, more than that and you need more parity data IMO. Even with ZFS, I am paranoid.
|
|
#
?
Mar 27, 2012 23:20
|
|