On Sunday 12 January 2014 15:56:32 Andrius D. Ilgunas did opine: > -- > Andrius > > > On Sun, Jan 12, 2014 at 1:21 AM, Gene Heskett <ghesk...@wdtv.com> wrote: > > so when 500Gig and up hard drives became available, I converted to > vtapes > > > on a hard drive, which has turned out to be, dollar for dollar, about > > 100x more dependable, > > Awesome, that answers another question I had - "Why use tapes these > days?" > > > Linux has excellent drive health drive monitoring utilities, so the 3 > > drive failures I have had since converting about 8 or 9 years ago, > > has always given me sufficient warning that no data has been lost, > > not a byte. > > Yes it does. Though I'm running a mostly FreeBSD shop and I just > recently started using zfs primarily because of its mirroring/data > correction capabilities, which I find very valuable in these days of > >1TB HDDs. > > Speaking of which, I see that amanda has a amzfs-sendrecv function, > which I presume is a wrapper around getting zfs snapshots stored > somewhere. But I'm not yet clear on how it (zfs/sendrecv) "fits" in > amanda's backup scheme. Would it simply take snapshots and save them > on a (v)tape? And when it comes time to recover, I would guess that > amrecover would rebuild the fs by restoring snapshot deltas? Sounds > like some more reading in this facet is needed. > > > Then about 12 years back I realized that a bare metal recovery to a > > newly formatted operating drive would always be a 1 days loss because > > the amanda data would also be lost, what was on the tapes was a day > > old, I wrote some wrapper scripts that make a tarball each of the > > instant configuration directory, and a tarball of the data directory > > amanda keeps and appends them both to the tape just used, so that > > when doing the full recovery, it started from the last known good > > state. > > Good one! I'll keep that in mind when building the backup system. > > > Unfortunately, even with the binlog enabled, some of the queries > > don't > > > > > get recorded (long story), > > > > And one that needs attended to in case forensics are required. > > Totally. It's just one of those problems inherited from the previous > DBA/SA > > > > Another part of my backup plan is going to have two configs. One > > > for the database instance ('hypercritical data') that will be > > > backed up insanely often, and another config for merely really > > > important data (svn repos, wikis, etc) which will have a less > > > taxing schedule. > > > > Separate configs need separate storage, you cannot mix tapes if you > > want to keep your sanity. Let each amanda config keep its own > > database and its own named tapes. Its really quite good at it. > > I'm not quite clear on your response. Creating a separate storage > volume isn't a problem since I'm using vtapes. Do you mean that it > would be easier to have one config to rule them all? Where "them" are > the different priority files and "them" are different clients backing > up to the backup server.
If separate configs, which I can't personally find an overpowering reason for, you would need, most likely, two separate tape "libraries" each containing its own drive(s), or 2 separate big hard drives. With the tapes labeled specifically for each use so that if somebody puts the wrong tape in the libraries changer, it would refuse to use it, thereby safe keeping the data. I think that same idea, where you have large data dumps, could easily be extended to HD's. Now, here is a thought, and I have no clue about zfs, I have never used it, but Tiger Direct has a 3 drive quick change cage that holds 3 3.5" drives in the same rack space as 2 half height 5.25 drives. No other special hardware, you just power down, open the retainer door in front of the drive, it slides out about 1/2", you finish pulling it out and slide another drive into the slot, close the door & power up. Sells for about a 70 dollar bill, and surprise, the commodity 1 Tb seagate also sells for about a 70 dollar bill. $280 gives your 3Tb of quickly removable storage. I am sure you could find a way to make that simplify things. A warning since linux has had a propensity to move drives around from sda to sdd or wherever, I think depending on the timing of the drive ready discovery process, that you will NOT want to mount drives as /dev/sd(abcdef) or as UUID=big number statements but as a LABEL=name drive is labeled as type of mount statements in the /etc/fstab file. That way you could have a dozen drives labeled as DailyOS##, and another dozen labeled as DailyDB## and with some advice from Jean-Louis, figure out a way to have a years worth of 30 day backups no farther away than a compartmented file drawer. It would take all that much effort to come up with a divider design that could hold those 2 dozen drives in one desk drawer, or better yet, a slide out on HD drawer sliders tray in a fireproof safe. Another thought about those drives is that they will probably need a firmware update before being put in service, and the Seagate site makes getting the right file to put on a bootable cd an absolute piece of cake. And while Seagate probably doesn't recommend it, I have updated a drive with an install on it without any loss, but the drive was 2 to 4x faster after the the flash. FWIW, but its a bigger secret, those LTO drives can probably benefit from a firmware update too. But the tape people keep that info well hidden. One of the other disadvantages to tapes is that drive parts are factory installed only. I am a C.E.T. and can replace a head in a dvc-pro broadcast vcr, and have it all calibrated in about 2 hours. The tape people will not sell me a replacement head drum for a DDS drive for any price. The dvc-pro drum and motor costs about $3500, easily 10x a precisely made, and about midway in size between a $30 VHS drum and the dvc-pro drum which a US quarter dollar about covers the top of. And after several years of trying to maintain those drives, it became glaringly obvious that they really really wanted to spend the holidays from Thanksgiving to New Years, in Oklahoma City. An emergency HD replacement is as far away as your nearest Staples. That alone is worth quite a bit. > > Or would it make more sense to have separate configs, e.g. > CriticalDatabasesOnHost1_config, ImportantFilesOnHost1_config, > ImportantFilesOnHost2_config, MiscFilesOnHost3_config, etc ? > > > Thats my $0.02, but adjust for inflation since 1934. Yeah, I'm an > > older user. ;-) > > Thanks so much for your input Gene! And, that just goes to show how much > value can be found in $0.02 from an older user. :) Your 2 pennies are > worth $4.94 !! Chuckle, so it would appear that my opinion is "appreciating" ;-) Too bad my SS check doesn't grow at that rate. I've been paying into that ponzi scheme since I was about 14 & drawing on it for about 14, I had it started some before I retired & almost worked till I was 68. But I was having back problems which I am about to go get an estimate on. Finally. :( Cheers, Gene -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Genes Web page <http://geneslinuxbox.net:6309/gene> Please, Mother! I'd rather do it myself! A pen in the hand of this president is far more dangerous than 200 million guns in the hands of law-abiding citizens.