Re: online 7.22 onarchive questions
In article <331F7A3E.11EB@titan.com>, Shimon Gonen <sgonen@titan.com> writes >I have read TFM, but can someone give me a few simple answers. > >The database is continuously updated (16MB logical log data per hour). > >Need the ability to revert the database to a previous state, and the >ability to restore the database up to last valid transaction. > >The database gets archived (to tape) in the following manner: > >day 1 - level 0 >day 2 - level 1 >day 3 - level 1 >day 4 - level 1 >day 5 - level 0 >... > >continuous logical log backups to disk (16MB per hour). >When restore is required - always do full (all dbspaces) restores. > >1. How do you restore (revert) a database to the state at a given level >1 backup (say day 3) ? a) Restore previous level 0 archives. (i.e. day 1) b) Restore latest level 1 archive since previous level 0 archive. i.e. day 2 c) Restore latest level 2 archive since previous level 1 archive (none). d) Rollforward all logical logs since last archive restored (level 0/1/2) i.e. all logical logs since level 1 archive on day 2. > >2. What logical logs do you need in order to do a full restore after a >crash on day 4, >2 hours after level 1 archive? >Do you need all logical logs since last archive level 0, or are the >logical logs since the last level 1 archive enough. > Since last level 1 is enough (see above). >3. Do you need all level 1 archives to restore to the day 4 level or >does the day 4 archive include all the info in the previous level 1 >archives? > You need the latest level 1 archive since previous level 0 archive. i.e. the day 4 archive. The slowest part (apart from full level 0 archive) is the rollforward of the logical logs. Response from Informix UK was to consider level 2 archive during the middle of the day. ALSO NOTE: When you do a restore the first thing you are prompted for is a new logical log tape to backup the logical logs which are on disk but not on tape. i.e. have a spare tape ready in case you need to bo a restore. When restoring I normally a) Get all users out of the system b) Use tbtape -a to flush all available logical logs to tape. c) Take a level-0 archive before restoreing. This means if the restore fails I have the option of restoring using c) and patching it i.e. I can choose the best version either the last archive or c) and work from there. I'd rather have the option of c) still open to me. Also b) may be used in a restore to allow restore up to the 'latest' version. It is very unlikely that online will rollforward a corrupt logical log since it does consistency checking when rolling forward. If this 'latest' tape fails to restore then just redo the restore excluding this latest tape. >Thanks. > -- David Williams