Restoring failed Informix mirror chunk - informix

Question

What should be done if an Informix mirror chunk failed and needs to be replaced with new chunk?

Answer 1

One command that you could use is onspaces. Run by a DBSA with no arguments, the help output includes the information:
onspaces -m <spacename> { -p <path> -o <offset> -m <path> <offset> [-y] |
-f <filename> }
onspaces -r <spacename> [-y]
onspaces -s <spacename> -p <path> -o <offset> {-O | -D} [-y]
-m — Add mirroring to an existing DBspace, PLOGspace, BLOBspace or
SBLOBspace
-r — Turn mirroring off for a DBspace, PLOGspace, BLOBspace or SBLOBspace
-s — Change the status of a chunk
Managing mirroring for a complete dbspace
Clearly, you could turn mirroring off for the down chunk (with -r) and then turn it back on with -m. You should investigate the Administrator's Reference — especially the section on ON-Spaces — and maybe the Administrator's Guide too.
The ON-Spaces section on mirroring contains the note:
The mirrored chunks should be on a different disk. You must mirror all the chunks at the same time.
and the syntax diagram allows multiple occurrences of the -p <path> -o <offset> -m <path> <offset> part of the synopsis. The -p and -o portions identify the existing chunk, and the -m portion identifies the new mirror chunk.
The -f option allows you to put the per-chunk information into a text file.
This technique has the not necessarily desirable side-effect of dropping all mirroring on the affected dbspace temporarily, and then reinstating it.
Managing single mirror chunks
Further research reveals a way to recover a single chunk at a time. The Administrator's Guide has a section on:
Fault tolerance
with sub-sections:
Mirroring
Using mirroring
These cover the theory and practice of mirroring. In particular, you seem to need:
Take down a mirror chunk using onspaces
onspaces -s db_acct -p /dev/mirror_chk1 -o 0 -D
Recover a mirror chunk using onspaces
onspaces -s db_acct -p /dev/mirror_chk1 -o 0 -O
This allows you to specify the chunk that is down — if the system has not already marked it down. And you can then bring it back online (into recovery mode) when you've replaced the physical media. As noted in a comment, this is much easier when you use symlinks to name the device (file) that holds the data (and if you don't use non-zero offsets; for the most part, they're a relic from the days when big disk drives were 100 MiB or less).