We've noticed exactly the same thing after upgrading to a kernel 2.4.20-18.7 one of our redhat servers. The kscand is not really documented. I've found in the Redhat kernel sources (linux/mm/vmscan.c) the kscand function :

/*
* The background page scanning daemon, started as a kernel thread
* from the init process.
*
* This is the part that background scans the active list to find
* pages that are referenced and increases their age score.
* It is important that this scan rate is not proportional to vm pressure
* per se otherwise cpu usage becomes unbounded. On the other hand, if there's
* no VM pressure at all it shouldn't age stuff either otherwise everything
* ends up at the maximum age.
*/
int kscand(void *unused) ...


This code is present into the official kernel sources ... I found the associate patch in the kernel workgroups (http://groups.google.com/groups?q=kscand+patch). No concrete information about the kscand daemon and its behaviour ... I didn't find any documentation into the redhat kernel documentation :o(

Is this CPU time consommation is normal ? It's our last redhat development server .. if I doesn't have a quick response .. I migrate it under a debian like all other servers. The provided kernel doesn't used undocumented code.

Alban


-- redhat-list mailing list unsubscribe mailto:[EMAIL PROTECTED] https://www.redhat.com/mailman/listinfo/redhat-list

Reply via email to