gshaw Posted June 9, 2011 Posted June 9, 2011 Just wondering what devices people are using for their primary disk-based backups? Up until now we've been doing full backups to tape from our physical servers but after moving to virtualised servers we're aiming to move to Veeam instead. Thing is we don't have a lot of space in the building to put an additional server (our office is pretty much the only place outside the server room) so can't have anything too bulky or noisy. However we do need to make sure the backup data is solid so would need some decent quality drives... any recommendations? Was thinking possibly QNAP, Netgear Readynas or something along those lines but can I trust a box like that for important backup data?
dhicks Posted June 9, 2011 Posted June 9, 2011 Thing is we don't have a lot of space in the building to put an additional server (our office is pretty much the only place outside the server room) so can't have anything too bulky or noisy. We use a QNAP server for backups here at school with 2TB WD Caviar Green drives. I use the same model of harddrive at home on my always-on server in our living room - they are very quiet (not absolutly silent, but pretty much) and energy efficient. The QNAP server itself seems to be running a passivly-cooled Atom processor of some sort, so again is very quiet. 1
gshaw Posted June 9, 2011 Author Posted June 9, 2011 We use a QNAP server for backups here at school with 2TB WD Caviar Green drives. I use the same model of harddrive at home on my always-on server in our living room - they are very quiet (not absolutly silent, but pretty much) and energy efficient. The QNAP server itself seems to be running a passivly-cooled Atom processor of some sort, so again is very quiet. Is that for your only backups or do you have it replicated somewhere as well? How about tape archive, do you still do that as well as the backup to QNAP?
dhicks Posted June 9, 2011 Posted June 9, 2011 Is that for your only backups or do you have it replicated somewhere as well? No replication, those are our only backups. How about tape archive, do you still do that as well as the backup to QNAP? No.
jamesfed Posted June 9, 2011 Posted June 9, 2011 We've got a QNAP TS-210 with 2x2TB WD Green drives for our Primary School that we support - only the one server so we run Windows Server backups overnight and connect to the drive using iSCSI. No tape backup at all as the QNAP is on an isolated portion of the site so its the whole unless a 747 lans on top of Abingdon were fine....
gshaw Posted June 9, 2011 Author Posted June 9, 2011 I usually use a meteorite landing on the building as my worst-case scenario although a 747 would work pretty well too
PeteHopton Posted June 9, 2011 Posted June 9, 2011 Don't use Caviar Green Drives in Linux based devices! There is a conflict between the head parking 'green' feature 'intellipark' and the way in which linux runs RAID, it means your heads will park and unpark every 8 seconds and wear out very quickly. Use RE4 disks or Seagate Constellation ES disk. If you get a windows based server with these disks in, Starwind do a virtual tape drive software that will work with all your existing and legacy backup systems. 2
morganw Posted June 9, 2011 Posted June 9, 2011 I've got two 2TB Caviar Green drives in a software RAID 1 array in an HP Microserver. I haven't noticed any delays or noises that makes me think that it's doing anything bizarre. Is there a good way to monitor the amount of head parks?
morganw Posted June 9, 2011 Posted June 9, 2011 This site seems pretty clued up about it: wdantiparkd - WD anti-intellipark daemon According to his formula (Take Load_Cycle_Count and divide it by Power_On_Hours and if the value is > 50, your drive is affected), mine doesn't seem to be. Mine is 12920 divided by 1650, which gives me a value of 7.83. Maybe they fixed it?
PeteHopton Posted June 9, 2011 Posted June 9, 2011 That is still pretty high, just not insane. Check the other half of the Raid mirror, the way Linux software RAID works means one disk is worse off. At that rate the disk will kill itself in 2-3years with the unit continuously on. There are other problems with 'caviar green' disks in raid or multiple disk configurations, a vulnerability to vibrations and an issue with how the drive reports a read error or bad sector to the RAID, which would be a recoverable operation from an RE disk but not a caviar green.
gshaw Posted June 9, 2011 Author Posted June 9, 2011 I wouldn't go with Green drives anyway, usually go for Samsung drives as a personal choice. The issue I have is whether to have a physical Veeam server or use their virtual appliance mode. The virtual appliance is good in that it's quicker to backup VMs stored on NFS but will be backing up over the network. The physical box would backup to local drives but will probably end up having to read files over the network due to slow read speeds on NFS datastores. You could argue after the first backup (big one of everything) subsequent backups are "forever incremental" so network load shouldn't be much of an issue. Only thing would be if I want to archive to tape the backup would have to be network based and that could be a lot of data (800GB max for the capacity of the tapes) Too many decisions!!!
morganw Posted June 9, 2011 Posted June 9, 2011 That is still pretty high, just not insane. Check the other half of the Raid mirror, the way Linux software RAID works means one disk is worse off. The second drive gives 12983 as a Load Cycle Count and 1652 Power On Hours. Dividing the first by the second gives 7.86 so it seems that both drives are doing the same thing. There definitely isn't any clicking either.
morganw Posted June 9, 2011 Posted June 9, 2011 I've decided to take both drives to work tomorrow and I'll try to use th wdidle3 tool to increase the time before parking. I forgot that I had a third one in there as well, it's older but it's only parked about 200 times. Thanks for the heads-up on this.
PeteHopton Posted June 10, 2011 Posted June 10, 2011 Backups Are really important things, the last thing you want to happen is a few random read failures on restore and the backup fail. My tips are this: Use enterprise grade HDD only. The rate of read failure is usually one order of magnitude lower and the RAID command set will allow recovery of bad sectors 'on the fly' reallocating spare sectors and rebuilding the data from the RAID set. Low cost high capacity disks include RE4 from WD and Constellation from Seagate. Use a decent RAID card with a battery backup unit. Why? lets say you server fails due to a power and UPS failure. Was the disk based backup forced off too? if so, and it was writing data without a BBU, then the RAID can become punctured and your array will fail on restoring the server. Use a double parity RAID on all arrays greater than a couple of TB. RAID 6 can save your data, these days, a non-recoverable read error is actually quite a common occurrence, every 100 TB or so read and there will be a NR read error. Let's say your backup is 10 TB usable and in RAID 5, a single disk fails, The RAID card now has to computer through about 14 TB of read data to rebuild the array, one non recoverable read error and the operation fails. Do you fancy your chances 14%? 1
enjay Posted June 10, 2011 Posted June 10, 2011 Using a Buffalo Terastation here - cost was a few hundred quid, very small, works a treat. Integrates to AD, so permissions are nice and easy to set.
enjay Posted June 10, 2011 Posted June 10, 2011 Can you use Windows ACLs on it? Was that question at me? If so, the answer is that I don't know, as I've never tried - controlling access rights by AD groups has been sufficient for everything I've needed to do (restrict access to the backup share to just IT Support, and create a second share for photos which staff can modify but students only read).
dhicks Posted August 4, 2011 Posted August 4, 2011 Don't use Caviar Green Drives in Linux based devices! There is a conflict between the head parking 'green' feature 'intellipark' and the way in which linux runs RAID, it means your heads will park and unpark every 8 seconds and wear out very quickly. Just bringing this slightly older thread back to life again - I've been looking at RAID issues again to day as we had a disk failure (caused by a blunt, heavy object crashing in to our server rack cage rather than anything more subtle or technical). Researching a bit, I came accross this: Why RAID 6 stops working in 2019 | ZDNet ...which points out that those WD disks have an unrecoverable read error rate (URE) of 10^15, rather than the more normal value for "non-enterprise" drives of 10^14. "Enterprise" drives generally have URE rates of 10^16, but that still gives the WD drives an order of magnitude greater difference in URE rates. The auto-disk-parking issues actually looks easy to solve, i.e. just make sure you write something to the disk every 8 seconds, although I don't know how that affects the WD drives' "green" credentials - are those drives inherently more energy-efficient when running than other drives, or do all drives draw about the same amount of power when doing reads/writes and rely on things like auto-parking to make them "green"? -- David Hicks 1
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now