ron
Member
We have two Sun servers: V480 for Production and E3500 as Backup. The E3500 is a warm spare - and is updated about hourly using AI files. A tape backup is taken each evening from the Backup server so that Production is not interrupted. We have nine DBs -- the largest is 126 GB.
We're using Progress 9.1D06.
All this works very well -- however I'm toying with a change the plan so that the nightly backup goes to disc first ... and then to tape. The idea being that the backup to disc should be faster than the backup to tape (over 5 hours) and so the updating of the warm spare database would not have to be suspended for such a long time. (And there are a couple of other advantages to be gained as well.)
HOWEVER ... I'm having trouble getting this to work as I would like. I will explain what I've done and what I'm trying to achieve ... and maybe there's a guru at large who can help me.
CASE #1
probkup to <disc>
tar or cpio the <disc> file to <tape>
This arrangement is not satisfactory. It does solve the problem of getting a "fast" backup to disc in the first place ... but whenever a restore from tape is required it is a two-stage process:
tar or cpio the <tape> file to <disc>
prorest the <disc> file
That would take too long and would need a second area on disc to restore from tape (because I would not necessarily want to over-write the most recent backup on disc).
CASE #2
probkup to <disc>
dd if=<disc> of=<tape>
This is much more satisfactory - because what is on the tape can be immediately restored using prorest. Unfortunately, however, the dd takes an unbelieveable length of time ... and so does the prorest! Clearly the problem here is that dd writes a very large number of very small blocks to tape (512 bytes each, I think).
I tried to overcome this by giving the dd an obs=8192 (or larger) block size. This improved the time to write to tape by a large factor ... but the resultant tape could not be restored by prorest. In every case the result was:
SYSTEM ERROR: Read of /dev/rmt/0 failed, errno = 12 (1052)
SYSTEM ERROR: Error reading backup header (1057)
!!! ERROR - Database restore utility FAILED !!! (8564)
So ... what can I do? Can anyone suggest a way to achieve what I want, namely:
I would be most grateful for any help.
Ron.
We're using Progress 9.1D06.
All this works very well -- however I'm toying with a change the plan so that the nightly backup goes to disc first ... and then to tape. The idea being that the backup to disc should be faster than the backup to tape (over 5 hours) and so the updating of the warm spare database would not have to be suspended for such a long time. (And there are a couple of other advantages to be gained as well.)
HOWEVER ... I'm having trouble getting this to work as I would like. I will explain what I've done and what I'm trying to achieve ... and maybe there's a guru at large who can help me.
CASE #1
probkup to <disc>
tar or cpio the <disc> file to <tape>
tar or cpio the <tape> file to <disc>
prorest the <disc> file
CASE #2
probkup to <disc>
dd if=<disc> of=<tape>
I tried to overcome this by giving the dd an obs=8192 (or larger) block size. This improved the time to write to tape by a large factor ... but the resultant tape could not be restored by prorest. In every case the result was:
SYSTEM ERROR: Read of /dev/rmt/0 failed, errno = 12 (1052)
SYSTEM ERROR: Error reading backup header (1057)
!!! ERROR - Database restore utility FAILED !!! (8564)
So ... what can I do? Can anyone suggest a way to achieve what I want, namely:
Initially backup to disc with prorest.
Copy the disc backup file to tape (in a similar time-frame to the time probkup would write a backup to tape).
Be able to prorest the tape file directly - without having to restore to a flat file first.
I would be most grateful for any help.
Ron.
