Restoring a level-1 backup at a later date?
Malc (IDS 9.30, HP-UX 11i) wanted to migrate a 180GB instance to new hardware with minimal downtime: take a level-0 ON-Bar backup days ahead, restore it to the new box, then later apply a level-1 to catch up, cutting outage from ~19 to ~5 hours. Doubts arose because the pending-level-1 info lives in the ixbar file, so the level-0 restore may just go to fast recovery. Suggestions: test the sequence on a small instance, use logical logs, or set up HDR for near-zero downtime; Neil Truby advised also taking an ontape backup rather than trusting Data Protector/ON-Bar. No confirmed outcome is recorded — the poster was testing on a small instance and leaning toward a physical restore followed later by a logical restore.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
Here's a good one. (IDS9.30, HP-UX 11i) We're moving to a different facility and new hardware (same basic machine & OS) at the end of the month and in order to minimise the time taken to transfer our 180Gb instance and allow the users unfettered access up to the very last minute (god we're so good to them) we'd like to do the following using OnBar: 1) Run a level-0 backup on the Thursday night from the old production server (currently this takes about 7 hours). 2) Restore this to the new server at the new facility (L0 restore time at present is about 12 hours end-to-end). 3) Leave the new server in limbo while the users have continued access to the old box. 4) Friday night kick the users and then take a level-1 of the old production box - we've tested this and it takes about 2.5 hours. 5) Take this level 1 and restore it to the new box (which should take about 3 hours) and let the users onto the new box. Thus the users should lose only 3 hours instead of 12. Can we do this? From the OnBar guide it looks like we can't but our scenario isn't covered. Sorry, we haven't had time to do a test run so we're after list wisdom (anyone called Obnoxio please pipe down!) Malc (Copying to ids@iiug.org)
Actually, I buggered up the timing calculations <fx:chagrin>. What I meant is that if we do a level-0 on the Friday night after the last user logs off then restore this to the new box this means the users are off for 7 hours for the backup and then 12 hours for the restore to the new box, i.e. 19 hours minimum. If we can do the deferred level-1 thing, that means they're off for 2.5 hours for the level-1 backup and 3 for the restore or only 5 hours in total.
Sounds OK, but you would have to level the new server in "recovery mode" for the few days having told it you have got a level one archive you want to restore. i understand you don;t have time to test your whole production instance, but you could maybe create a new small instance on a machine, build the stores database perform the level 0 and recover it on the other server. Then perform some activity on the first server and take a level 1 archive. restore this on the second server and then bring it online. See the changes you made are there. At least that way you have performed the steps, probably about an hour or 2s work, and spotted any glaring problems. Are you using Logical Logs? If so you could even allow the users to continue working even after the level 1 has completed, apply that then kick the users off and replay the logical logs on the new server. Or how about you set up HDR on Thursday, replicate everything until the weekend, then kick the users off the old server, restart the new server in Standard mode and you should only have minutes of downtime instead of hours. malc_p@btinternet.com wrote: > Here's a good one. > (IDS9.30, HP-UX 11i) > We're moving to a different facility and new hardware (same basic > machine & OS) at the end of the month and in order to minimise the time > taken to transfer our 180Gb instance and allow the users unfettered > access up to the very last minute (god we're so good to them) we'd like > to do the following using OnBar: > 1) Run a level-0 backup on the Thursday night from the old production > server (currently this takes about 7 hours). > 2) Restore this to the new server at the new facility (L0 restore time > at present is about 12 hours end-to-end). > 3) Leave the new server in limbo while the users have continued access > to the old box. > 4) Friday night kick the users and then take a level-1 of the old > production box - we've tested this and it takes about 2.5 hours. > 5) Take this level 1 and restore it to the new box (which should take > about 3 hours) and let the users onto the new box. Thus the users > should lose only 3 hours instead of 12. > > Can we do this? From the OnBar guide it looks like we can't but our > scenario isn't covered. > Sorry, we haven't had time to do a test run so we're after list wisdom > (anyone called Obnoxio please pipe down!) > Malc > (Copying to ids@iiug.org)
Sounds OK, but you would have to level the new server in "recovery mode" for the few days having told it you have got a level one archive you want to restore. i understand you don;t have time to test your whole production instance, but you could maybe create a new small instance on a machine, build the stores database perform the level 0 and recover it on the other server. Then perform some activity on the first server and take a level 1 archive. restore this on the second server and then bring it online. See the changes you made are there. At least that way you have performed the steps, probably about an hour or 2s work, and spotted any glaring problems. Are you using Logical Logs? If so you could even allow the users to continue working even after the level 1 has completed, apply that then kick the users off and replay the logical logs on the new server. Or how about you set up HDR on Thursday, replicate everything until the weekend, then kick the users off the old server, restart the new server in Standard mode and you should only have minutes of downtime instead of hours. malc_p@btinternet.com wrote: > Here's a good one. > (IDS9.30, HP-UX 11i) > We're moving to a different facility and new hardware (same basic > machine & OS) at the end of the month and in order to minimise the time > taken to transfer our 180Gb instance and allow the users unfettered > access up to the very last minute (god we're so good to them) we'd like > to do the following using OnBar: > 1) Run a level-0 backup on the Thursday night from the old production > server (currently this takes about 7 hours). > 2) Restore this to the new server at the new facility (L0 restore time > at present is about 12 hours end-to-end). > 3) Leave the new server in limbo while the users have continued access > to the old box. > 4) Friday night kick the users and then take a level-1 of the old > production box - we've tested this and it takes about 2.5 hours. > 5) Take this level 1 and restore it to the new box (which should take > about 3 hours) and let the users onto the new box. Thus the users > should lose only 3 hours instead of 12. > > Can we do this? From the OnBar guide it looks like we can't but our > scenario isn't covered. > Sorry, we haven't had time to do a test run so we're after list wisdom > (anyone called Obnoxio please pipe down!) > Malc > (Copying to ids@iiug.org)
Not sure we can tell it there's a level-1 to come - that's held in the ixbar file isn't it? So the level-0 would just exit to fast recovery without asking for a level-1 unless we can hack the ixbar......... ugh.
Could you not do a test on a smaller database server?
btw - Mr Paranoid-but-rightly-so-IMHO here - If I were you I would NOT rely
upon Data Protector and onbar to happen to deceide to work together on the
day, and would take an ontape backup too!
--
Neil Truby t:01932 724027
Director m:07798 811708
Ardenta Limited e:neil.truby@ardenta.com
<malc_p@btinternet.com> wrote in message
news:1107425675.546800.92530@z14g2000cwz.googlegroups.com...
> Here's a good one.
> (IDS9.30, HP-UX 11i)
> We're moving to a different facility and new hardware (same basic
> machine & OS) at the end of the month and in order to minimise the time
> taken to transfer our 180Gb instance and allow the users unfettered
> access up to the very last minute (god we're so good to them) we'd like
> to do the following using OnBar:
> 1) Run a level-0 backup on the Thursday night from the old production
> server (currently this takes about 7 hours).
> 2) Restore this to the new server at the new facility (L0 restore time
> at present is about 12 hours end-to-end).
> 3) Leave the new server in limbo while the users have continued access
> to the old box.
> 4) Friday night kick the users and then take a level-1 of the old
> production box - we've tested this and it takes about 2.5 hours.
> 5) Take this level 1 and restore it to the new box (which should take
> about 3 hours) and let the users onto the new box. Thus the users
> should lose only 3 hours instead of 12.
>
> Can we do this? From the OnBar guide it looks like we can't but our
> scenario isn't covered.
> Sorry, we haven't had time to do a test run so we're after list wisdom
> (anyone called Obnoxio please pipe down!)
> Malc
> (Copying to ids@iiug.org)
>
<malc_p@btinternet.com> wrote in message
news:1107429166.336614.266270@l41g2000cwc.googlegroups.com...
> Not sure we can tell it there's a level-1 to come - that's held in the
> ixbar file isn't it? So the level-0 would just exit to fast recovery
> without asking for a level-1 unless we can hack the ixbar......... ugh.
ontape! ontape!
Yup we're giving it a bash on a smaller instance - looks like using a physical restore followed at a later date by a logical restore might be the option. And I haven't the foogiest clue as to why you are so cynical about HP's flagship DataProtect product being even slightly unreliable. Shame on you. ;-) Malc