Re: hung process

In article <5lfnqs$4uf@cssun.mathcs.emory.edu>, Oscar Goldes <ogolde@impsat1.com.ar> writes >Hello, >I am using informix online 7.2 over SCO Unix. >I frequently notice that there are a lot of hung process. >Specifically, I saw that two process were eating up nearly 100 % cpu time (a >pentium biprocessor) as seen with vmstat or cpusar. >Both of them had huge ammounts of cpu time as by ps. >The two processes turned out to be dbaccess sessions, which had 1 as parent >process, but did not belong to root, just to normal users. I saw this as well on a Sun Solaris 2.5.1 box running OWS 7.20.UC1. I had two of them keep the PU 0% idle. One of them had been running for two weeks and had accumulated around 20000 minutes of CPU time!!. onstat -u shows the sessions were not reading /writing and onstat -g ses..even showed one of them have no current database! I suspect users turning off PC's or closing terminal sessions without quiting dbaccess. Theere was a known 4GL bug which sounds similar to this..we the process 'lost' it's controlling terminal it would enter a tight loop continually reading 0 bytes from the terminal i.e. no error checking was done on the read() call from the terminal... >They could be seen from onstat also. >Killing them solved the cpu usage problem. However, such process tend to appear >frequently. >Any suggestions? > >TIA > >Oscar Goldes > -- David Williams