Re: Lockups!

In article <38momh$aa4@emory.mathcs.emory.edu> informix@asl3.asl-labs.bc.ca (Informix Sysadmin) writes: >We've recently been enduring a spate of lockups. > >Scenario...The sqlturbo process starts up and goes NOWHERE. > >any user on the database gets complete freeze. > >The usual check of tbstat -l shows that we have lots of log space, but >that a checkpoint is required. > >An attempt at tbmode -c goes off for a while, and then returns the >prompt, having NOT done a checkpoint, but certainly NOT told us about any >error. > >"Force checkpoint" from the online menu locks tbmonitor menu up solid. > >We can take the database offline with an "Immediate shutdown", but this >is not really an acceptable solution. Anyone have any ideas why this is >happening? Do you need more info? If so, I can gather at the next >occurence. > >Brian P. Hampson > Do a tbstat -o savefile1, tbstat -A > savefile1.a, and call support. You have a buffer that is held by one process which is necessary for the engine to complete the checkpointing process. Questions to ask are - is the process functioning ? - is something rolling back a long transaction ? I've seen this twice when on an unstable matrix of operating system software. A process was hanging in the kernel, and never came out. Normally it was aggravating, but when the process was an application frontend and the backend had an active buffer, it was fatal. The problem can be either an o/s bug, an informix bug, or bad coding compounded by operator error. Informix support is the place to start for this one.