I guess nobody uses telnet anymore
... because it appears that there's a huge honking bug in login(1) on recent versions (recent being "since r*dh*t 8.0 was released) of Linux, where it doesn't bother to recycle old utmp entries, and thus utmp -- the file that records the people currently logged in -- grows without bound when a systems integrator who uses telnet (and thus login(1)) does a performance test of telnet sessions.
Questions like "why is /var/run/utmp 10mb long?" are not likely to bring out the best in me, particularly since utmp has been around for AT LEAST 30 YEARS NOW and there's no reason why utmp handling should mysteriously break inside login. I look at glibc, and mutter darkly about Unix going malignant, and mutter more because, if my suspicions are correct, I'm going to have to pester r*dh*t, the maintainers of util-linux, and whoever the hell maintains gl*bc these days to accept patches that fix this wonderful behavior.
Comments
Don’t forget the other aspect of this - why would RedHat care about a file growing to 10 megabytes when “everybody” is using 500Gb Logical Volumes? It’s only dinosaurs like me who carefully size our partitions.
Comments are closed
What you’re seeing there is a side effect of the massive deprecation of telnet in favour of ssh; I would not be surprised if Redhat considered using telnet a firing offense.