Re: Database Replication - Does it work?

Kenneth McCarthy wrote: > > Has anyone been able to automate informix Database replication recovery? > > I've been trying now starting with R6.0, R7.10 and now R7.12. using two PPC/AIX > servers connected via a LAN. > The data is short lived - few hours to days, but the data is all critical to > the real time applications writing and reading the data. > > The idea of using DR was that the data is replicated on a second > server providing a hot standby backup. The applications and the > environment that they run in are well behaved and invarient, and are required > to provide 7/24 hour availability. The system must also be maintenance free > (or as much as possible - if h/w fails, manual recovery will need to be performed). > > After starting the secondary from a level 0 archive of the primary, > the paired machines run okay with data replicated from the primary to > the secondary - all client applications are connected to the primary. > Because I don't need to backup logical log data > (my backup is the other machine), I have LTAPEDEV pointing to /dev/null on both > primary and secondary. > > If the primary fails (onmode -k), the secondary changes mode to standard, with > replication off - my client applications detect the failed primary, and reconnect > to the secondary now standard - everything is still as it should be. > > If I recover the primary oninit before the secondary log files start > overwritting the the primary recovers - Great. > > If the logfiles overwrite, I expected that when I try to recover the primary > that I would be prompted to recover first from tape backups and that I > would at this point just recover from scratch (archive 0 of the secondary) - okay. > HOWEVER, I am not given any indication that I need to recover from tapes: > The primary hangs in fast recovery and the secondary panics. At this point > to recover the system requires careful manual procedures - not good. > > I've tried to point LTAPEDEV to a real file and periodically backup logical > logs 'ontape -a' on the secondary server when it has changed to standard, but > when I try to restore the primary it hangs in fast recovery and the > secondary panics - again requiring careful manual procedures to recover. > (the idea with this approach is to write the logfiles to a file, then remove the file to > make the online believe that the data has been backed up and therefore will prompt > for a tape when I try to restart the primary). > > Has anyone been able to sucessfully automate DR - has anyone been able to use DR > sucessully any way? Informix Support keeps telling me that it works and that I must > be doing something wrong. > > Any hints/help would be appretiated. > > Ken McCarthy > 613-765-5226 I have DR running and have found a few problems with it, but none like yours. The first thing I would do is do everything by the book, so log your logical logs to tape by pointing ltapedev to a tape device and have continous backup running. See what happens. I haven't had the the chance to recover DR from from tape, but could it be that you don't have enough logical log space for the transactions and you hit the high water mark? Just a thought.