如何判断full recovery？

leetaedong

还有就是 不完全恢复后，log number是   会从 0000开始重置  还是 接着 最后利用的log number +1？

不完全恢复后，是不是一定需要重新备份？ 以前的备份会失效？

wangzhonnew

what exactly you mean "不完全恢复"?
restore and rollforward are totally separate concept. From the name "不完全恢复" it sounds like part of "restore" operation, or are you talking about something like "不完全前滚"?

and how do you define "不完全前滚"? from db2's perspective, once a database is restored, it doesn't know what is the current LSN from the pre-restored version of database, so there's no way to tell what is the expected LSN to be stopped. So when you fire rollforward command to end of logs, it will just search until hitting last valid transaction log and will stop. It's your responsibility (or any DBAs) to validate whether the transaction log is the latest one or not.

the next log extent id will not be reset to 0, instead it will start a new log chain and log extent id will be starting from the last log+1 from your rollforward operation.

leetaedong

我说的不完全恢复是 restore + rollforward to <time>（是oracle 里面的概念）。应该是你理解的" 不完全前滚“ 。

确实，rollforward 之后会 重新 开启一个 log chain。

xiaowg

原帖由 wangzhonnew 于 2011-8-11 20:53 发表
what exactly you mean "不完全恢复"?
restore and rollforward are totally separate concept. From the name "不完全恢复" it sounds like part of "restore" operation, or are you talking about something like "不完全前滚"?

and how do you define "不完全前滚"? from db2's perspective, once a database is restored, it doesn't know what is the current LSN from the pre-restored version of database, so there's no way to tell what is the expected LSN to be stopped. So when you fire rollforward command to end of logs, it will just search until hitting last valid transaction log and will stop. It's your responsibility (or any DBAs) to validate whether the transaction log is the latest one or not.

the next log extent id will not be reset to 0, instead it will start a new log chain and log extent id will be starting from the last log+1 from your rollforward operation.

你的回答和IBM的答案一样，我彻底服了。难道SAP不知道DB2的这个缺点？一旦关联的数据库需要恢复，整个SAP系统需要手工做数据校对，费时费力。

wangzhonnew

as long as you can make sure all log files are copied (by using mirror logging on separate file systems, you can guarantee the log files won't be missed unless someone manually remove them) you won't have integrity issues...

xiaowg

Since LUW DB2 does not know the last UOW, how can DBA know?

PS: Do copy primary logs before DB restoration, otherwise they will be removed.

wangzhonnew

once rollforward is done, it will show the timestamp where rollforward stops, and then you can see whether if it's the one around your outage time

xiaowg

how to get the exact outage time? guess? how to compare?

The IBM DB2 DBA told me that these 2 timestamps are not same.

I have done a test and found: even full recovery, these 2 timestamps are not same.

[ 本帖最后由 xiaowg 于 2011-8-29 13:15 编辑 ]

wangzhonnew

"how to get the exact outage time?"
there's something called "db2diag.log"

which 2 timestamps are not same?

wangzhonnew

leetaedong , i don't really get what exactly you are talking about, if you still have more concern please open a new post and show the reproduce script and screenshot to explain your concern.

Thanks