Re: Why this Long Transaction?

Joe Lumbley (jlumbley@netcom.com) wrote: > Folks, > Twelve hours ago my large database was shut down ungracefully, with > a monster insert operation going on. About 2 million rows in the table, > but I don't know how many were actually inserted. On restart, we went into > that misnomer of all misnomers , "fast recovery". That was about 11.5 > hours ago. It's been recovering since then. > At about the 10.5 hour mark, my tbstat -t output showed a Long Transaction > flag. My LTXHWM is 70% and my LTXEHWM is 80%. I'm currently at about 40% > log usage and nothing else is running. So why the long transaction? Are > there different rules for how the HWM's work in recovery? Is there some > kind of time check? I know it won't make any difference and I'll still be > recovering for another 45 minutes, but it's bothering my little monkey brain. I'm following up two days after the original post and hope you are not still in "fast recovery" but these symptoms bear a very strong resemblance to a deadlock I had with 5.0 about a year ago. My LTXHWM was 70% also and it just never came out of a fast recovery because I ran out of log space before the rollback completed. Did tbstat -l show that you were making progress on your logfiles? In my case I had to do a recover from the daily archive. I have since set LTXHWM/LTXEHWM to 45/50% respectively and no one will ever convince me to raise them. I consider the Informix defaults of 80/90 BTW to be downright dangerous. My $.02 contribution to your survival guide with Informix! Ken Miles Vancouver Division Hewlett-Packard