Backups to disk
Posted in 1994
Topics: Backup & Restore, Storage & Space Management, Logging & Checkpoints
Subject: Re: Back-ups to Disk? johnpo@vcd.hp.com (John Powell) writes: johnpo> This post is related to an earlier one about logical logs and named johnpo> pipes. I realize that Informix only supports archives and logical johnpo> logs going to tape (via tbtape on Rel. 5) at this time. However, I johnpo> would like the archive and logical logs to go to disk from which they johnpo> can be backed up to tape via the daily UNIX file system backups. I johnpo> imagine that if one were to dedicate enough disk space for this johnpo> purpose, write the appropriate shell scripts, and arrange for this area johnpo> of disk to be put to tape daily, then one would have a fairly reliable johnpo> backup strategy in place. Has anyone tried this yet? Has anyone johnpo> successfully restored (rolled forward) Informix from files on disk? johnpo> Please reply with any information you have on this subject. We've been doing this for more than a year now. It has been working without too many problems, but if something goes wrong it can be pretty bad. Here's how it works: Re-assign TAPEDEV to some file on disk. This file must already exist, and there must be enough space on available on disk for the entire dbspace, not just the data. Of course, ownership and permissions must be set correctly. As informix, run the script which performs the archive. This script does the following: force checkpoint force completion of current log determine pid of log archiver and kill it move log file to Old log file and create new log start log archiver program perform level 0 backup As Johnathon Leffler wrote, it can be bad if there is not enough disk available. Usually what happens is the file which contains the error log fills up with error messages asking for another tape. Around this time someone gets a call, because the machine is getting pretty bogged down. We go one further, though. After the archive is finished, we compress it using a zero-loss algorithm (Gnu Zip, or compress), and copy it across the network to another node offsite. We have successfully restored data from one of these files. Error checking at each step is pretty crucial (very crucial, but there are not enough tools to do it all--yet!) The code to do all this is pretty straightforward shell scripts. The heart of the matter is the redirection of the input to the archive process. If anyone cares to see the actual scripts, I can probably post them. _____________________________________________________ /| Clem Akins Standard Disclaimers Apply | || Reynolds Metals Co. ("Cutesy" ones, too!) | || Alloys Plant (MS #59) | || 3509 East 2nd St Internet: cwakins@rmc.com | || Muscle Shoals, Al Telephone: (205) 386-6783 | || USA 35661-1256 Facsimile: (205) 386-6831 | || "Climb High, Cave Deep!" | ||_____________________________________________________| /_____________________________________________________/