-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Kern Sibbald wrote: > On Thursday 20 April 2006 19:14, Ryan Novosielski wrote: >> Kern Sibbald wrote: >>> On Thursday 20 April 2006 18:26, Ryan Novosielski wrote: >>>> Hi all, >>>> >>>> After a loss of my director machine earlier this week (someone "thought" >>>> the rootdisks were mirrored -- no such luck), I have decided that >>>> bacula_rescue is probably the way to go. We had no UFS dump or anything >>>> like that, so we just happened to be lucky that an older copy of Solaris >>>> 9 was still on the box somewhere. The box is the director for my Bacula >>>> backups, but the machine itself is backed up by our datacentre, which >>>> uses HP's Data Protector. I'm not sure whether I will endeavor to change >>>> this or not, but at the very least the rescue could get me going either >>>> way I decide to go. >>>> >>>> I see that there is a make_static_bacula script in the directory, but it >>>> occurred to me that this is not going to be enough if I don't have a >>>> director to use for the restore. Can someone point me in the right >>>> direction for my attempt to do this? Would I just modify the scripts to >>>> build the whole package static so that I have it for my restore? >>>> >>>> Thanks for any assistance that you can provide. >>> To give an appropriate answer, you will need to be a bit more specific >>> about what was knocked out in your loss (i.e. DIR, SD, database, ...) and >>> exactly how you want to recover. On how you want to recover, I mean: on >>> the same machine with a new harddisk, elsewhere, ... >> I lost the entire root volume, which contained essentially everything on >> the box. The machine was backed up by Data Protector, so I was able to >> boot the other copy of Solaris 9 and have data protector >> remote-installed and be on my way. What I lost was an install of Bacula >> that had the DIR, SD, and FD. Therefore if I had a rescue that only >> included the FD, I'd still be up the creek. Probably my most likely plan >> is to backup the base OS with Bacula just to get to a restore point (far >> enough up that I can have Data Protector installed) and then worry about >> restoring the rest from the datacentre backup. The situation with this >> box is that it is a webserver as well as Bacula director/sd/fd. I am >> concerned with my ability to quickly return the machine to service to >> continue making backups -- the other mission of the box is not my >> responsibility. The root volume is now mirrored and this SHOULD never >> happen again, but we know how that goes. >> >>> IMO, the simplest and quickest way to get a Bacula server back up is >>> first to have a Bacula rescue disk, or perhaps the new Knoppix remastered >>> disk that includes the Bacula rescue files, AND a separate machine where >>> you can at least temporarily bring up a Director, SD, and a database. >>> With the DIR, SD and a database on a separate machine, you can then >>> repair or recreate your hardisk partitioning, load up a static FD, and >>> restore everything ... If I lost my Bacula server that is what I would >>> do. I always have a "hot spare" with a SCSI card, and my Bacula conf >>> files waiting for the time when my backup server totally dies -- >>> hopefully this will never happen. >> This is a Sun UltraSparc E450, not a Linux machine. Does Knoppix exist >> for UltraSparc? > > No I don't imagine so. > >> Actually, is the "solaris" tree in the rescue perhaps >> aimed toward Solaris x86? In that case, I suppose I'm out of luck anyway. > > No, it was aimed at Sparc, but no one ever offered to make it work -- I don't > have a Sparc and the one I had access to I never had root permissions so > could never run the scripts anyway ... > >>> In most companies, the cost of having a "hot spare" is about $300 for a >>> second SCSI card, and maximum 100 MB of disk space holding Bacula and a >>> dummy catalog database. At the time of a problem you will need >>> sufficient extra disk space to handle the full Bacula catalog. >> The machine is much more hardened now and everything is mirrored, with >> hot spares in addition. I also have 3 SCSI controllers (a total of 6 >> channels) so I'm set in the case of any kind of large-scale failure, but >> I still need to be able to restore this machine in the case of a >> disaster of a larger scale. I likely will not have extra hardware for a >> backup director, so I'd like to find out how I can make a complete >> enough rescue disk/disc to get back on my feet. Perhaps some builtin >> Solaris utility would be a better idea, then restoring Bacula from Data >> Protector as I did this time. However, my datacentre is relatively slow >> with restores and this time I missed 3 days of backups by the time the >> box was back in shape. > > Well, I haven't looked at the Solaris rescue code for a long time, so I doubt > very much that it is at all useful. It would be rather a lot of work to get > it into shape. > > My previous recommendation still goes, have a "standby spare" on a separate > Solaris system ready to go. If you don't have a spare Solaris system, the > next best bet is to spend about $1500 for a suitable Linux machine that could > at least host the Dir, FD, and database to restore your other machine. This > would not be so simple though since it is not clear that restoring a database > written on a Solaris to a Linux machine would work. In addition, there would > be some rather important changes needed to your SD conf file because of drive > naming convention differences ... > > Now that you have had this problem, you may want to quickly glance over the > Rescue chapter in the manual, and more importantly read the Restore chapter > of the manual. The later has a lot of suggestions for how to recover in case > of disaster. Any comments about the doc in this area would be *much* > appreciated since there is nothing like someone who has just been through it > to know what the real problems are -- I can only guess ...
Perhaps at least part of what we can get out of this situation is to assist in the Solaris rescue portion of the package. I have a couple of Solaris boxes that are soon to be repurposed. As a rescue solution for boxes using Solaris would be very welcome (two of my -sd's are Solaris 10 machines that are as-yet unsupported by Data Protector which is Sparc only). I could test restores that resulted from the bare metal recovery, since the boxes are not in use at present anyway. I will take a better look at the docs later and let you know what kinds of things I ran into. It was a little different because I'm missing step one from my process (I did not need to get the machine booted as I found a slice with Solaris on it and did all of my work from that environment on a mounted set of disks elsewhere in the chassis). While bacula didn't help me in this restore as it wasn't being used, I at least appreciate the fact that once it was restored from tape, it kept on humming right as it did when I left off. :) - -- ---- _ _ _ _ ___ _ _ _ |Y#| | | |\/| | \ |\ | | |Ryan Novosielski - User Support Spec. III |$&| |__| | | |__/ | \| _| |[EMAIL PROTECTED] - 973/972.0922 (2-0922) \__/ Univ. of Med. and Dent.|IST/AST - NJMS Medical Science Bldg - C630 -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.2.2 (MingW32) iD8DBQFER82Qmb+gadEcsb4RAqecAKCyz9fWMZCI7dYZNtXQ9EYOjwhyGACfZEKh tiQxMtCl/CunCaPAb1SvNcs= =jdld -----END PGP SIGNATURE----- ------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
