[linux-elitists] A better cron: large production environments
matt at deploylinux.net
Mon May 10 15:59:13 PDT 2010
> Handling concurrency is getting to be important for
> things like this.
Yes -- concurrency management is imho the biggest issue with existing cron implementations. Having dozens or hundreds of linux systems all attempting to wake up and run cron.daily roughly at around the same time and performing disk intensive io/etc and hitting the same SAN/etc is no longer an uncommon event.
I've seen multiple kludges to work around this, and I notice that fedora 12 seems to have more variation in cron start times than RHEL 5 -- but I'd still like to have an easy to install cron jobs across a large number of servers and not so much specify when they start as much as the maximum number that can run at once and the particular time window in which they should run.
(e.g. in a cluster of 100 VM's, be able to quickly state that daily cron jobs should run between 4-5am, but no more than 5 VM's should execute their jobs at once).
Matthew Marlowe - http://www.deploylinux.net/
CEO @ DeployLinux Consulting, Inc
matt at deploylinux.net, 858-400-7430
More information about the linux-elitists