Hi, On Sunday 06 July 2008 23:46, Petter Reinholdtsen wrote: > The idea is that one should add the hostname of the host in question > (FQDN) in the fsautoresize-hosts netgroup, and this should be enough > to activate the cron job on that host. Is this a good way to do it? Sounds like to me. > The script will syslog every time it (try to) extend a file system, so > the changes will not happen in complete silence, but we should > probably try to get a nagios warning sent as well to allow the events > to be reported to those that want to know about them. I suspect it > will also trigger an email from cron when it happen, but want to quite > down the script to make sure that do not happen unless some unexpected > error happen. I definitly want to see a syslog entry for when it runs (with success or not), too, so that it shows up in logcheck too. > I'm not sure how often the resizing should be done, but propose once > per hour to have reasonable response time when a disk is filling up. Not sure how invasive the check is, but maybe even every 15min? > The debian-edu-fsautoresize script uses the rules in > debian-edu-fsautoresizetab when deciding how to handle the different > file systems, and the current rules look like this. > > # Example configuration for fsautoresize on Debian Edu. > # Override these values in /etc/fsautoresizetab > # > # regex minfree max extendby > /.* 10% 20g 10% > /usr 10% 10g defaults Today (as opposed to some years ago) I dont think /usr should be on a seperate partition anymore. I know about the benefits of having partitions per service, but with todays terrabyte drives that has become a lot less useful and more partitions mean more cognitive stress for system administrators. > /var 10% 10g defaults > /var/spool/squid 10% 40g defaults is squid configured to use that much space? the default is 256mb... regards, Holger
Attachment:
pgpkR_JGoQtTL.pgp
Description: PGP signature