continuous log backup to disk (ontape -c) ... again
Posted in 1998
Glenn,
We have not implemented anything in this area yet, but have considered the
problem of thinking about processes that tend to be oriented to tape
on disk. Except for data replication as offsite backup protection, we
still do our stuff ontape.
What about running periodic ontape -a processes and renaming
the files after they are done?
For example 8:00 am:
cron runs a shell:
that runs ontape -a to file log_log_bu_out
and renames log_log_bu_out to log_log_bu_file1
10:00 am
ontape -a to log_log_bu_out
and renames log_log_bu_out to log_log_bu_file2 etc
You assure that other processes are backing up log_log_bu_file1 etc to
tape or offsite media.
I think the down side is a period of time during which you are not backing
up logical logs and you need to allocate a fair amount of disk. Of course
you can always salvage logical logs if you need to restore, assuming
your catastrophe doesn't involve the drive holding the log backups or you can
run ontape -a more frequently to minimize the risk. Simply run every 10
minutes instead of 2 hours as suggested above and adjust the tape or
off-site backup accordingly.
Easy to say when I'm not doing it, but it seems like a possibility.
Good luck
Con Woodall
Colorado State University Veterinary Teaching Hospital
cwoodall@vth.colostate.edu
970-491-1244
On Thu, 2 Apr 1998, Glenn Travis wrote:
> I am wanting to run continuous log backups to disk and have been successful
> in laying out the initital strategy;
>
> . full backup (ontape -s -L 0) nightly
> . then start a script to the log backup in the background;
> ontape -c <<EOF >ontape.c.log
> ^M
> EOF> . kill the ontape -c script before running the next backup.
> . backup the log file (LTAPEDEV). (move or rename or delete, then touch a
> new one).
> . restart script (ontape -c) in background.
>
> This works fine EXCEPT when the LTAPESIZE is reached by the background
> ontape -c process.> Since this process is running unattended in the background (we passed the
> RETURN ^M character to initiate), the next prompt to mount a new LTAPEDEV
> (in this case a filename), spins out of control, filling the ontape.c.log
> logfile. (I use the output in this logfile to get the log file numbers
> which were backed up). If the size of LTAPEDEV is reached before the
> process is killed prior to the nightly backup, the scripts loops, asking for
> a new file, eventually filling up disk space.
>
> I know you are going to suggest making LTAPEDEV large enough to hold all the
> logs for the day. Well, 2gb is the limit and sometimes we reach that. So,
> I want it to wait until I back it up and touch a new one. I don't want it
> to keep asking for a new file. Why does it think I've entered a return to
> the prompt, when I only passed one return character to the ontape -c
> command?
>
> any ideas on solving this one?
>
> thanks,
> Glenn Travis
> Sr. DBA
> Circuit City / Divx
>
> Informix 7.23uc9
> Sun solaris 2.5
>
>
>