Hard Crash – Clean of the site

[Tony Benitz]

Version 5.2 on a Sun box

We had a hard crash (lost a CPU) on our production interface box.  We have a High Availablity software that allowed for a fail over.  There were issues with get the processes back up and I was forced to do a manual clean of the each site.  I have a script that says that it cleans the site, but it only removes the shmem file and the pid files.

My questions have to do with some changes that I want to make to my site clean script.

I am planning on adding

– the removal of the sem* files from the exec directory

– hcidbinit -i (initialization of the ICL database)

– keybuild rlog

– dchain rlog

My question is .. Am I at risk of losing any messages.

🙄

Topic

[Bill Bertera]

I just wrote a similar script with the commands you just mention. And I tested it with messages in the recovery database, and they were still there when I brought the site back up. You should be fine. This is what I ran:

hcisitecleanup

rm $HCISITEDIR/exec/databases/vista.taf

rm $HCISITEDIR/lock/*

hcimsiutil -Z

hcidbinit -if

keybuild rlog

keybuild elog

dchain rlog

dchain elog

rm $HCISITEDIR/exec/sem*

[Rob Abbott]

Make sure you do a “hcidbinit -i” at the very least – the standard HACMP start script does this – or you will have routing problems.

The threads register themselves in the ICL database with an IP and port.  If the IP changes the ICL database needs to be re-initiailized.

Rob Abbott
Cloverleaf Emeritus

[Author]

Replies