Hello, On Mon, 2008-06-09 at 07:33 +0200, Petter Reinholdtsen wrote: > [Franklin PIAT] > > Yep, it's less intrusive, however, "ionice -p" doesn't seems to be > > propagated to existing child process : > > I know, and that is OK, as long as it is executed early in the script, > before child processes are forked.
So you did meant "ionice -c2 -p $$" before invocation, not "ionice -c2 -p $!" after... I miss-understood that. As you mentioned, it would be less intrusive, so it would be better. > > The class 3 (idle) can lead to deadlocks[1]. (I think there were some > > improvements in recent kernels). > > Ah, good to know. > > >> I thought class 2 was the default, thus making the operation a no-op? > > You are right... I usually run "nice ionice -c2 foobar" (which works). > > Maybe we should do that... or use "ionice -c2 -n7" > > If -c2 already is the default, it would be a no-op, and I see be no > reason to call it for the popcon scripts. Class 2 is the default class. But within that class there are many priorities. The manpage says "This class takes a priority argument from 0-7, with lower number being higher priority". ioprio.txt[1] states that "io_nice = (cpu_nice + 20) / 5" which means the default is 4. > The '-n7' argument might have some effect, but with -c2 it would raise the > priority of the > process, not lowering it, if I understand ionice correctly. Yet > again, I fail to understand why you would like to use the arguments > you propose. Care to elaborate? You pointed out my mistake previously : "ionice -c2" should be run in conjunction with "nice", or "ionice -c2 -nX" to actually have an effect. > Bill suggested this should be done in cron, and not in each cronjob. > What is your view on this? Why should it be fixed here and not in > crond? I've just replied to his mail. Short answer : "Why not if it can be in Lenny". Long answer include : "...undesirables side effects..." Franklin -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
