On Sat, Feb 6, 2010 at 7:11 PM, Darren Reed <Darren.Reed at sun.com> wrote: > Mike Gerdts wrote: > ... >>> >>> Perhaps ZFS would never actually write >>> these files, so maybe this wouldn't be an issue with ZFS though. From >>> what I've seen, /var/tmp typically has a lot of small files in it anyway, >>> many of them zero length. >> >> And on the other end of the spectrum are files that are huge that have >> been sitting there for years with no one ever cleaning them up. ?I >> think that there is a compelling case for making /var/tmp its own file >> system (rpool/var/tmp) with a quota on it (to limit overall size) or >> perhaps user quotas (to discourage irresponsible use). ?Those that >> argue for a separate /var to get noexec,nodevices may like to set >> those properties on /var/tmp and not worry so much about a separate >> /var. > > These are nice ideas that reflect the new spaces into > which this can grow but I suspect that a quota is going > too far for a default image.
Agreed. I didn't intend to say that it belongs as a default, but rather an option. > We've all spent the last N years doing single partition > installs because disks grew to the point where that was > the best thing to do (unless you really wanted 100G of > space disk in /export/home.) > > I have one question for you and this proposal from Sanjay: > > What happens when an application stores peristent data > under a shared directory in /var and there are multiple BEs > that have their own version of this application and the > binary data format is not compatible? <sarcasm>You write a postinstall script that does the conversion as the relevant package is installed.</sarcasm> More seriously, I'm not really sure. I think that a snapshot is important to be able to provide fallback for when things go wrong and for when some sort of a data conversion is required and fallback is needed. This should not be as encumbered as a boot environment snapshot/clone. Hopefully the occurrences where such conversions are needed are kept to a minimum. Does data exist to suggest that this is a frequent occurrence? Does it typically happen with data in /var or data that is more likely to be kept in a pool other than rpool? There are plenty of other cases where there is potential harm with the current scheme. Imagine checks queued in the printing system that get printed during the image update and after the subsequent boot. Or the answer to a marriage proposal that is queued but never deliverd. (OK, that is contrived - anyone that would do a marriage proposal through email is very unlikely to have someone to propose to...) Perhaps less serious would be forgotten DHCP leases or lost log entries. -- Mike Gerdts http://mgerdts.blogspot.com/
