Hey folks,
I just went through the archives to see what people are doing for backups, and here is what I found : - amanda - bacula - BackupPC - FreeNAS
Here is my situation : we have pretty much all Sun hardware with a Sun StorageTek SL24 tape unit backing it all up. OSes are a combination of RHEL and CentOS. The software we are using is EMC
NetWorker Management Console version 3.5.1.Build.269 based on NetWorker version 7.5.1.Build.269
The pickle we are in right now is that this software is Java based, and stops working at a very specific release of JRE (1.6.26 or something like that). We still have some machines around with that release and it looks like we need to keep at least 1 of them, but this is clearly not a long term viable solution.
In the end I want to get our central IT group to take over our backups if possible (we are a bit of an island outside of central IT), but as I pursue that path I also want to pursue a 2ndary path assuming they will say "no".
I am familiar with BackupPC and will look at the other recommendations above. I think that Bacula and Amanda are sort of the drop-in replacements for what we have now so I'll look at them most closely. But if I do have to carry forward with our own backups I'd ideally like to get out of the tape game - never liked tapes.
Anyway, since the last big backup discussion was over a year ago I figured I'd kick off another one to see if anything new has come up in the mean time.
What are the current recommendations?
cheers, -Alan
NetWorker Management Console version 3.5.1.Build.269 based on NetWorker version 7.5.1.Build.269
The pickle we are in right now is that this software is Java based, and stops working at a very specific release of JRE (1.6.26 or something like that). We still have some machines around with that release and it looks like we need to keep at least 1 of them, but this is clearly not a long term viable solution.
I'm pretty sure I saw a note on the networker list that 7.6 SP3 works with update 27, update 29, and java 7.
On Thu, Dec 8, 2011 at 8:53 AM, Alan McKay alan.mckay@gmail.com wrote:
Here is my situation : we have pretty much all Sun hardware with a Sun StorageTek SL24 tape unit backing it all up. OSes are a combination of RHEL and CentOS. The software we are using is EMC
NetWorker Management Console version 3.5.1.Build.269 based on NetWorker version 7.5.1.Build.269
The pickle we are in right now is that this software is Java based, and stops working at a very specific release of JRE (1.6.26 or something like that).
That sounds like something that can/should be fixed.
I am familiar with BackupPC and will look at the other recommendations above. I think that Bacula and Amanda are sort of the drop-in replacements for what we have now so I'll look at them most closely. But if I do have to carry forward with our own backups I'd ideally like to get out of the tape game - never liked tapes.
If you want mostly-online backups with perhaps an occasional tar archive, it will be hard to beat backuppc because of it's storage pooling and ability to run over rsync or smb with no remote agents. For all-tape, I'd probably go with amanda because of its ability juggle the full/incremental mix automatically to fit the available tape size. I haven't used bacula but it looks like it might be good if you want a mix of online and tape storage and can deal with the agent installs.
On Dec 8, 2011, at 8:43 AM, Les Mikesell wrote:
On Thu, Dec 8, 2011 at 8:53 AM, Alan McKay alan.mckay@gmail.com wrote:
Here is my situation : we have pretty much all Sun hardware with a Sun StorageTek SL24 tape unit backing it all up. OSes are a combination of RHEL and CentOS. The software we are using is EMC
NetWorker Management Console version 3.5.1.Build.269 based on NetWorker version 7.5.1.Build.269
The pickle we are in right now is that this software is Java based, and stops working at a very specific release of JRE (1.6.26 or something like that).
That sounds like something that can/should be fixed.
I am familiar with BackupPC and will look at the other recommendations above. I think that Bacula and Amanda are sort of the drop-in replacements for what we have now so I'll look at them most closely. But if I do have to carry forward with our own backups I'd ideally like to get out of the tape game - never liked tapes.
If you want mostly-online backups with perhaps an occasional tar archive, it will be hard to beat backuppc because of it's storage pooling and ability to run over rsync or smb with no remote agents. For all-tape, I'd probably go with amanda because of its ability juggle the full/incremental mix automatically to fit the available tape size. I haven't used bacula but it looks like it might be good if you want a mix of online and tape storage and can deal with the agent installs.
---- also - Bacula now has 'Enterprise' version with SLA and yes, Bacula can not only do tape and/or disk but can also migrate backup jobs (ie, disk to tape)
Craig
Le jeu 08 déc 2011 09:43:21 CET, Les Mikesell a écrit:
On Thu, Dec 8, 2011 at 8:53 AM, Alan McKay alan.mckay@gmail.com wrote:
Here is my situation : we have pretty much all Sun hardware with a Sun StorageTek SL24 tape unit backing it all up. OSes are a combination of RHEL and CentOS. The software we are using is EMC
NetWorker Management Console version 3.5.1.Build.269 based on NetWorker version 7.5.1.Build.269
The pickle we are in right now is that this software is Java based, and stops working at a very specific release of JRE (1.6.26 or something like that).
That sounds like something that can/should be fixed.
I am familiar with BackupPC and will look at the other recommendations above. I think that Bacula and Amanda are sort of the drop-in replacements for what we have now so I'll look at them most closely. But if I do have to carry forward with our own backups I'd ideally like to get out of the tape game - never liked tapes.
If you want mostly-online backups with perhaps an occasional tar archive, it will be hard to beat backuppc because of it's storage pooling and ability to run over rsync or smb with no remote agents. For all-tape, I'd probably go with amanda because of its ability juggle the full/incremental mix automatically to fit the available tape size. I haven't used bacula but it looks like it might be good if you want a mix of online and tape storage and can deal with the agent installs.
In this last scenario, dar (http://dar.linux.free.fr/doc/Features.html) works just fine and don't need any remote agent. It is also at least as fast as Bacula at restore time, provided the "catalogue" is ready.
On Thu, Dec 8, 2011 at 9:55 AM, Philippe Naudin philippe.naudin@supagro.inra.fr wrote:
If you want mostly-online backups with perhaps an occasional tar archive, it will be hard to beat backuppc because of it's storage pooling and ability to run over rsync or smb with no remote agents. For all-tape, I'd probably go with amanda because of its ability juggle the full/incremental mix automatically to fit the available tape size. I haven't used bacula but it looks like it might be good if you want a mix of online and tape storage and can deal with the agent installs.
In this last scenario, dar (http://dar.linux.free.fr/doc/Features.html) works just fine and don't need any remote agent. It is also at least as fast as Bacula at restore time, provided the "catalogue" is ready.
That looks like a one-off kind of tool. Backuppc/amanda/backula are all frameworks to manage potentially large numbers of targets.
Another interesting thing is Relax and Recover (http://rear.sourceforge.net/ - in EPEL as rear). This is something that you run on a working system to generate a bootable iso with that system's own tools to reconstruct the current filesystem layout (including LVM/md raid, etc.) and restore a backup onto it. It includes a few backup methods internally but with a small amount of work you could integrate your own backup approach into it to get a fully-scripted bare metal restore.
On Thu, Dec 8, 2011 at 10:31 AM, m.roth@5-cent.us wrote:
I just went through the archives to see what people are doing for backups, and here is what I found :
- amanda
- bacula
- BackupPC
- FreeNAS
You missed rsync.
Rsync is another one-off approach where you have to roll your own commands per target. Backuppc can use rsync as the transport, collating all the results into one centrally managed archive with a web interface that makes it easier to set up than rsync itself. Plus it will compress the data and pool all identical content so you can keep much more history online than you would expect.
Les Mikesell wrote:
On Thu, Dec 8, 2011 at 10:31 AM, m.roth@5-cent.us wrote:
I just went through the archives to see what people are doing for backups, and here is what I found :
- amanda
- bacula
- BackupPC
- FreeNAS
You missed rsync.
Rsync is another one-off approach where you have to roll your own commands per target. Backuppc can use rsync as the transport,
<snip> Actually, my manager wrote a set of scripts some years ago, and we *do* have centralized backup setups, which get automagically pushed out, and the backup hosts know what directories to backup from each server.
But it is a roll-your-own, though I'd have to go look to see if he released it as FOSS.
mark
On Thu, Dec 8, 2011 at 11:00 AM, m.roth@5-cent.us wrote:
I just went through the archives to see what people are doing for backups, and here is what I found :
- amanda
- bacula
- BackupPC
- FreeNAS
You missed rsync.
Rsync is another one-off approach where you have to roll your own commands per target. Backuppc can use rsync as the transport,
<snip> Actually, my manager wrote a set of scripts some years ago, and we *do* have centralized backup setups, which get automagically pushed out, and the backup hosts know what directories to backup from each server.
But it is a roll-your-own, though I'd have to go look to see if he released it as FOSS.
But is it better somehow than backuppc, which is basically a perl script that: (a) can use rsync, tar, smb, or ftp to collect the backups (b) provides a web interface with the ability to delegate host 'owners' (c) schedules everything for you (d) optionally compresses (e) detects and pools files with duplicate content, even from different sources. (f) is packaged in EPEL
It does have its own quirks, of course. The main ones being that its rsync-in-perl (on the server side so it can work with its own compressed files while chatting with a stock remote rsync) is somewhat slow, and that its archive storage that uses hardlinks for pooling may end up being impractical to copy with file-oriented tools. But basically it just takes care of itself after the initial setup.
Anyone have any experience with this, which just came to my attention
http://www.arkeia.com/en/solutions/open-source-solutions
--On Thursday, December 08, 2011 01:06:10 PM -0500 Alan McKay alan.mckay@gmail.com wrote:
Anyone have any experience with this, which just came to my attention
Yes, I've used it (albiet about 8 years back or so), as well as many other solutions (both commercial and FOSS) over the years.
Being bit by Arkeia (and previously Amanda and others) is why I started to design my own FOSS solution at the time ... however I was about 6 months into the design when Bacula first appeared, written by Kern with whom I had worked on a different project. (Actually, it was scary ... between my design and what he had, there was only about one significant difference at the time.)
So I shelved my project and started using bacula, and haven't looked back.
Use bacula. Drink the coolaid. Be happy.
BTW, my observation with a lot of the products at the time was that they *mostly* worked, but when the edge cases caused me to lose backup history, or made it impossible to back up certain filesystems, or had other unusual problems, it made it obvious that something better was needed. Without naming names, there was one FOSS product I evaluated that, when I looked at the source, they weren't doing any decent level of error checking ... which would explain the SEGVs I was seeing.
Once bitten, twice shy.
I will allow for the possibility that Amanda and Arkeia may have improved over the years, however Bacula was already a solid product when they were flakey or had other limitations, and has just gotten better with time.
Devin
On Fri, Dec 9, 2011 at 11:44 AM, Devin Reade gdr@gno.org wrote:
Being bit by Arkeia (and previously Amanda and others)
Errr, what? Amanda is a little cumbersome to set up, but it doesn't bite. If gnutar works, amanda should work or tell you why.
--On Friday, December 09, 2011 11:48:49 AM -0600 Les Mikesell lesmikesell@gmail.com wrote:
Errr, what? Amanda is a little cumbersome to set up, but it doesn't bite. If gnutar works, amanda should work or tell you why.
As I said, it's been at least 8 years since I dealt with Amanda. Going by memory, though, in the case of Amanda, it wasn't flakiness but rather limitations.
At the time Amanda had the limitation that the backup of a filesystem could not span tapes. This was a critical issue in that I had filesystems larger than the largest tapes available. In the case where filesystem sizes approached the size of a tape, it wasted a lot of tape space (which wasn't cheap).
I'm willing to believe that they've fixed that limitation, but if so I'd already moved on.
Arkeia, though, was definitely one of the flakey ones (IIRC I lost three months' worth of backups, which was caught by a scheduled validation process ... luckily it was only history and not current data that was gone.)
Devin
On Fri, Dec 9, 2011 at 12:11 PM, Devin Reade gdr@gno.org wrote:
--On Friday, December 09, 2011 11:48:49 AM -0600 Les Mikesell lesmikesell@gmail.com wrote:
Errr, what? Amanda is a little cumbersome to set up, but it doesn't bite. If gnutar works, amanda should work or tell you why.
As I said, it's been at least 8 years since I dealt with Amanda. Going by memory, though, in the case of Amanda, it wasn't flakiness but rather limitations.
At the time Amanda had the limitation that the backup of a filesystem could not span tapes. This was a critical issue in that I had filesystems larger than the largest tapes available. In the case where filesystem sizes approached the size of a tape, it wasted a lot of tape space (which wasn't cheap).
The brilliant thing about amanda, going back more than a decade, is that it knows how to estimate the size of backups and if you give it many filesystems to back up, it will skew a mix of full and incremental runs to fit the tape efficiently, getting at least an incremental every day and as many fulls as will fit. Of course you will cause problems if you put more data than will fit on your tape on a single filesystem, though.
I'm willing to believe that they've fixed that limitation, but if so I'd already moved on.
I think they have, but I just let my old system run until the tape drive died and by than was more than satisfied with backuppc, using a raid-mirroring scheme to make offsite copies (soon to be replaced independently running offsite servers). I'll be happy if I never see a tape again.
--- Les Mikesell lesmikesell@gmail.com
--On Friday, December 09, 2011 01:03:18 PM -0600 Les Mikesell lesmikesell@gmail.com wrote:
I'll be happy if I never see a tape again.
Likewise.
Skipping forward to the present, I'm doing normal backups in Bacula to virtual volumes on hard disk. As for offsite/archival backups, using the bacula add-on 'vchanger' and inexpensive high density SATA disks on an eSATA peripheral to act as a virtual tape autoloader magazine is both faster and less expensive than tape.
Devin
On Fri, Dec 9, 2011 at 1:52 PM, Devin Reade gdr@gno.org wrote:
--On Friday, December 09, 2011 01:03:18 PM -0600 Les Mikesell lesmikesell@gmail.com wrote:
I'll be happy if I never see a tape again.
Likewise.
Skipping forward to the present, I'm doing normal backups in Bacula to virtual volumes on hard disk. As for offsite/archival backups, using the bacula add-on 'vchanger' and inexpensive high density SATA disks on an eSATA peripheral to act as a virtual tape autoloader magazine is both faster and less expensive than tape.
Bacula is probably better suited to mixing online/tape or fake-tape for offsite, but you can't keep as much online as backuppc without help from ZFS or similar block-level dedup. I doubt if it can match the bandwidth efficiency of backuppc with rsync as the transport (not sure - how does the bacula agent deal with growing files, or big files with small changes?).
--On Friday, December 09, 2011 02:59:08 PM -0600 Les Mikesell lesmikesell@gmail.com wrote:
I doubt if it can match the bandwidth efficiency of backuppc with rsync as the transport (not sure - how does the bacula agent deal with growing files, or big files with small changes?).
There is a relatively new block-level "delta plugin" for that type of situation. I've not used it yet, so I don't have any opinions or comparisons.
I'm not sure if it's in the community edition yet (some functionality starts out in the enterprise edition and then later gets moved into the community edition).
Devin
Anyone have any experience with this, which just came to my attention
I have used Arkeia for a few customers .. it works well. Do you have any specific questions about it?
Barry
OK, I'm getting ready to finally dig into replacing our backups. Lots of good info in this thread -but so far no mention of rsnapshot
Any comment on it ? Our environment is all Linux except for Mac desktops which would like have a different solution for backups.
From the little I've read it seems to be very similar to BackupPC. Though
based on the name I guess it is using LVM snapshots? Which of course means almost instantaneous backups - attractive.
On Fri, Dec 16, 2011 at 12:34 PM, Alan McKay alan.mckay@gmail.com wrote:
OK, I'm getting ready to finally dig into replacing our backups. Lots of good info in this thread -but so far no mention of rsnapshot
Any comment on it ? Our environment is all Linux except for Mac desktops which would like have a different solution for backups.
From the little I've read it seems to be very similar to BackupPC. Though
based on the name I guess it is using LVM snapshots? Which of course means almost instantaneous backups - attractive.
Rsnapshot keeps what look like snapshots online with a history hardlinked where possible (in the history of the same system to save some space. I'm not quite sure how much it understands about lvm but it looks like that's a special case for local use.
Backuppc will give you compression as well, plus will hardlink all identical content even if found on different targets or in different locations, so you can keep much more history in the same amount of space. And it gives you a nice web interface to monitor and control everything.
good info in this thread -but so far no mention of rsnapshot
Any comment on it ?
we use rsnapshot. i think of it basically as a wrapper around rsync. it isn't a fully featured backup solution just on it's own, but it is a great tool. we have written a bash shell script wrapper around rsnapshot to do things like mount/umount the appropriate NFS storage brick to use as an rsync destination, and then send an email summary of the results if there was a problem.
From the little I've read it seems to be very similar to BackupPC.
I think the only thing they have in common is that they both use rsync as the transfer agent. Les describes how BackupPC has additional features like compression, hardlinking between different backup sets, and a web gui. I'd also add that BackupPC allows users to perform their own restore operations. rsnapshot has nothing like that.
Though based on the name I guess it is using LVM snapshots?
no, rsnapshot does not use LVM snapshots (at least, not that I know of). it used cp -al to create hardlinks between each backup run.
On Fri, Dec 16, 2011 at 5:17 PM, Jonathan Nilsson jnilsson@uci.edu wrote:
From the little I've read it seems to be very similar to BackupPC.
I think the only thing they have in common is that they both use rsync as the transfer agent.
Well, they are both perl scripts... Backuppc just has more of it.
Though based on the name I guess it is using LVM snapshots?
no, rsnapshot does not use LVM snapshots (at least, not that I know of). it used cp -al to create hardlinks between each backup run.
I think it can do something with them, but just locally so it isn't much of a backup.
On 12/16/2011 06:16 PM, Les Mikesell wrote:
On Fri, Dec 16, 2011 at 5:17 PM, Jonathan Nilsson jnilsson@uci.edu wrote:
From the little I've read it seems to be very similar to BackupPC.
I think the only thing they have in common is that they both use rsync as the transfer agent.
Well, they are both perl scripts... Backuppc just has more of it.
Though based on the name I guess it is using LVM snapshots?
no, rsnapshot does not use LVM snapshots (at least, not that I know of). it used cp -al to create hardlinks between each backup run.
I think it can do something with them, but just locally so it isn't much of a backup.
Does backupPC have the abilty to easily be configured so that each daily incremental and each weekly full backup are stored on different drives, i.e. to rotate drives based on your backup schedule and not just when a drive fills up? I think this might look something like having one drive for each day of the week, 1 for each week of the month, 1 for each month of the year, etc...
Thanks, Nataraj
On Sun, Dec 18, 2011 at 6:00 PM, Nataraj incoming-centos@rjl.com wrote:
Does backupPC have the abilty to easily be configured so that each daily incremental and each weekly full backup are stored on different drives, i.e. to rotate drives based on your backup schedule and not just when a drive fills up? I think this might look something like having one drive for each day of the week, 1 for each week of the month, 1 for each month of the year, etc...
No, due to the way that duplicate files are pooled with hardlinks, backuppc must store everything on a single filesystem and additional copies do not consume additional space - so there is no reason to do that from a storage perspective. However, from discussions on the mail list, I believe that there are people who swap/rotate the entire archive drive daily or weekly, letting it catch up after each swap, and others have various schemes to image-copy the filesystem with raid or lvm mirroring to get offsite copies.
So going back to Amanda and Bacula ... I seem to recall that Amanda uses standard tools on the back end like gtar and/or dump, is that right?
What does Bacula use? Does it use one of the standard tools? Or does it have its own proprietary format that it uses?
thanks, -Alan
On Mon, Dec 19, 2011 at 9:49 AM, Alan McKay alan.mckay@gmail.com wrote:
So going back to Amanda and Bacula ... I seem to recall that Amanda uses standard tools on the back end like gtar and/or dump, is that right?
Amanda needs its own client installation on the target, but uses the local gnutar or dump, then stores it with a header pre-pended so you can skip over it if you needed to restore with only standard tools.
What does Bacula use? Does it use one of the standard tools? Or does it have its own proprietary format that it uses?
Not sure about that - it does also have its own agent to install, though.
Les Mikesell wrote:
On Thu, Dec 8, 2011 at 10:31 AM, m.roth@5-cent.us wrote:
I just went through the archives to see what people are doing for backups, and here is what I found :
- amanda
- bacula
- BackupPC
- FreeNAS
You missed rsync.
Rsync is another one-off approach where you have to roll your own commands per target. Backuppc can use rsync as the transport, collating all the results into one centrally managed archive with a web interface that makes it easier to set up than rsync itself. Plus it will compress the data and pool all identical content so you can keep much more history online than you would expect.
I use backuppc, but find that in order to restore one has to be or know the admin user password. There appears to be no way to open this up to users to directly see and restore from the file tree that it manages.....
I use backuppc, but find that in order to restore one has to be or know the admin user password. There appears to be no way to open this up to users to directly see and restore from the file tree that it manages.....
Huh? No. Users can do their own restores from the web interface without root access. I think you need to go back and read the fine manual a bit more :-) There is definitely a way to set up users on there though. I have a fair bit of experience with BackupPC (great software)
On Thu, Dec 8, 2011 at 12:12 PM, Rob Kampen rkampen@kampensonline.com wrote:
I use backuppc, but find that in order to restore one has to be or know the admin user password. There appears to be no way to open this up to users to directly see and restore from the file tree that it manages.....
You can delegate target machines to 'owners' who can only see their own machines with their login to the web interface, but there is not an easy way to do it at the home directory or file owner level for a multi-user machine. You can make a 'host' which is a subset of a target, and point several of those at the same real host with the ClientNameAlias option but it would take some additional work to secure those against each other. It probably could be done, though.
On Thu, Dec 8, 2011 at 3:53 PM, Alan McKay alan.mckay@gmail.com wrote:
Hey folks,
I just went through the archives to see what people are doing for backups, and here is what I found :
- amanda
- bacula
- BackupPC
- FreeNAS
Here is my situation : we have pretty much all Sun hardware with a Sun StorageTek SL24 tape unit backing it all up. OSes are a combination of RHEL and CentOS. The software we are using is EMC
My non-tape solution of choice is definitely rsync => box with ZFS, snapshot however often you'd like. => forever incrementals.
For more redundancy and performance, add more ZFS boxes, do replication between them.
For tapes, I'd go with Bacula, but my intermediate storage will probably be ZFS anyway, for easy management of filesystems. I like creating one storage device per client as per this amazing write-up by Henrik Johansen: http://myunix.dk/category/bacula/
I'd choose Bacula mainly for experience and being comfortable with it. In this setup, I'm used to managing it all with Puppet:
From server to client to storage agents as well as creating individual
zfs filesystems for each client on the storage server. I had to patch the puppet zfs provider a while back to make it work on FreeBSD.
For Bacula, there now exists an awesome (modern) web interface, with ACL support and all: http://webacula.sourceforge.net/
Good luck. -- Mike
On 12/08/11 11:26 AM, Mikael Fridh wrote:
For more redundancy and performance, add more ZFS boxes, do replication between them.
what zfs replication is that? last I heard, the only supported replication was physical block replication of the underlying device(s) (avs in solaris cluster, drbd in linux), and the replica couldn't be mounted at all, it was purely for standby failover scenarios.
On Thu, Dec 8, 2011 at 8:38 PM, John R Pierce pierce@hogranch.com wrote:
On 12/08/11 11:26 AM, Mikael Fridh wrote:
For more redundancy and performance, add more ZFS boxes, do replication between them.
what zfs replication is that? last I heard, the only supported replication was physical block replication of the underlying device(s) (avs in solaris cluster, drbd in linux), and the replica couldn't be mounted at all, it was purely for standby failover scenarios.
What I mean is merely incremental zfs send -i | zfs receive -F between two boxes for each new snapshot being created. You're free to mount the filesystem, but each new receive will roll it back to the previous snapshot when another incremental comes in (using zfs receive -F).
It's not filesystem replication per se, but more periodic snapshots + incremental transfers. For doing multiple copies off backup data, I'd say it's more than good enough as "replication".
-- Mike