Re: Unix kill -9 and Informix SE???

In article <CnHzI9.57D@pinnacle.co.nz> jonc@pinnacle.co.nz (Jonathan Chen) writes: >It should issue SIGTERM (kill -15) to the processes first to request >orderly shutdown, wait a while, and if they're still alive, then only should >a SIGKILL (kill -9) be issued. I wouldn't do that (the kill -9) to an sqlexec. If the first signal doesn't kill it then it may be rolling back a transaction. Sending a SIGKILL during rollback will almost certainly result in a corrupt database, either internally or referentially. Personally I prefer to first kill the application with SIGTERM (15), SIGHUP (1), and finally SIGKILL if the others don't work (but they will), and then send a SIGTERM to the sqlexec and wait for it do die gracefully. >> (3) What are the possible side effects of using "kill -9" on a >> sqlexec process ( especially with the isql session in an >> update or insert state )? Corrupt btree indexes. Corrupt data pages. Partially committed transactions. Just say no. -- Hugh Grierson hugh@fujitsu.co.nz 22C, 70%rh, overcast