Date: Fri, 13 Feb 1998 09:32:01 -0500 From: sbabkin@dcn.att.com To: hackers@FreeBSD.ORG, matthew@netsol.net Subject: RE: Large system backups; recommendations for devices & strategie s? Message-ID: <C50B6FBA632FD111AF0F0000C0AD71EE413293@dcn71.dcn.att.com>
next in thread | raw e-mail | index | archive | help
[-- Attachment #1 --] The most I did was 6 systems (3 pairs) and 20G of data :-) The idea was the following: there is one backup system. Once a week a full backup is done manually (in meaning that it's necessary to manually insert tapes, kick off the backup scripts and check their results for errors, after that manually put tapes into safe). During the week this system has a tape with ongoing data. We changed this tape once a day (manually, with storing all the tapes unused at the moment in a fireproof safe) by two reasons: 1. In the worst case (like total destruction by fire) we can tolerate loss of one day of data, but better not more. 2. The tape is big enough (DDS-2 120m) to store all data for one day, so we don't need to change it more often. Application explicitly stores the data into a separate hierarchy of directories (the same as the main one). A particular case of that is Oracle archived logs. This second hierarchy gets polled, data transferred as cpio archive to backup server, saved in 2 copies. The backup servers restores this data into its main directory hierarchy, after what passes the cpio archive with server ID added to tape server and initiates removal of backed up data in the backup hierarchy on main server. After that backup server does any necessary additional things to newly received data, like applying Oracle archived logs. The tape servers pours everything it receives into the tape. The reason for explicit second hierarchy is an attempt to reduce overhead of searching for new data and reduce the period of polling. In case when restoration is needed, search the backup log for the latest occurrence of given file name, get its cpio archive name and date of copy, take the tape, restore archive from it (mt fsf is useful with it), and finally restore file from archive. And there was a like but separate technology for archived data that has to be stored for several years. Now I see that they have nothing like in AT&T :-) Although [mostly] batch processing of billing data has different requirements than online operation of a bank. -SB > ---------- > From: matthew@netsol.net[SMTP:matthew@netsol.net] > Sent: Thursday, February 12, 1998 8:05 PM > To: 'hackers@FreeBSD.ORG' > Subject: RE: Large system backups; recommendations for devices & > strategies? > > As many times as you can if all this data may e modified from minutes > ro minutes > matt at Future Lab > > ---------- > From: Mike Smith > Sent: Monday, February 09, 1998 7:56 PM > To: hackers@FreeBSD.ORG > Subject: Large system backups; recommendations for > devices & strategies? > > > (Please pardon the crosspost to -isp; I'm looking for comments > from > people with experience administering backup strategies for > largish > networks, and I suspect some of you lurk there.) > > I'm looking for recommendations for both backup devices and > backup > strategies for a network of about six systems and perhaps 50GB > of > data. Ultimately, I'd like something that can run more or less > unattended, modulo media changes, etc. (ie. I expect using > Amanda or > similar.) > > I'd be interested in hearing from anyone that's been involved in > > setting up and/or operating such a backup system, as well as > perhaps > being interested in doing something similar for the FreeBSD > project. > > -- > \\ Sometimes you're ahead, \\ Mike Smith > \\ sometimes you're behind. \\ mike@smith.net.au > \\ The race is long, and in the \\ msmith@freebsd.org > \\ end it's only with yourself. \\ msmith@cdrom.com > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe hackers" in the body of the message > > [-- Attachment #2 --] x>" IPM.Microsoft Mail.Note 1 ! D41F388339A4D111AF200000C0AD71EE D RE: Large system backups; recommendations for devices & strategies? ! &
