Question - Logical corruptions in REDO log group members

[JMac]

If the big advantage in having multiple members of REDO log groups on different devices is to safeguard against logical corruptions - as opposed to physical - then how does ARCH know which member to archive?

If I have 2 members in Group 1 and member 'a' has a logical corruption but member 'b' does not - when ARCH comes around to archive how does it know to use member 'b' and not 'a'?

[julian]

I don't think that ARCH checks for that but I am not 100% sure. Here is how the archiver process works:

ARCHIVER STEPS:

1) read the control file to find any unarchived logs,
2) open online redo log members to be read,
3) allocate redo log buffers (log_archive_buffers),
4) (async) read of the online redo log (log_archive_buffer_size),
usually aioread call if supported by operating system
uses alternative redo log members per buffer stream
5) fill redo log buffers,
6) (async) write of to the archive files (log_archive_buffer_size),
creates ufs archive file if not already created
first checks if buffer is full or if end of log
makes aiowrite call if supported by operating system
7) update control file with new information,
modify the archive log link list and change redo log statuses
8) starts the loop again.

[JMac]

Thanks Julian,

I was wondering just what the advantage is of having multiple group members if, as I do, the redo files are mirrored using RAID. I can see how RAID is good if you have a physical corruption but couldn't see how multiple members would help if the ARCH process didn't know which member to archive.

I might email support...

[JMac]

Thanks Julian,

I was wondering just what the advantage is of having multiple group members if, as I do, the redo files are mirrored using RAID. I can see how RAID is good if you have a physical corruption but couldn't see how multiple members would help if the ARCH process didn't know which member to archive.

I might email support...

[julian]

Originally posted by JMac
Thanks Julian,

I was wondering just what the advantage is of having multiple group members if, as I do, the redo files are mirrored using RAID.

OK, say you have only 1 member in group 3, and by accident you delete it with the RM command (for example). What do you do next?

Now, what if you had a second one. You just drop the deleted one from the DD, and recreate it.

[JMac]

If I did that Julian I'd eat my own feet.
Good argument - thanks.

[julian]

Originally posted by JMac
If I did that Julian I'd eat my own feet.
Good argument - thanks.

:-) I have done that once, lucky me, I had it multipxed :-)