Logs need to persist across reboots
Today there's no visibility if a user goes and reboots the box where we can trace back what happened to get the user into a potentially failed state.
/var/log should be a file backed memory disk (not a ram disk like it is today) that's synced periodically to a persistent store to ensure that we have traceability to determine cause of failure in user issues across reboots.
#1 Updated by Xin Li almost 8 years ago
/var/log is persist if user choose to configure it as such. The problem - it's not supported to have /var on a ZFS so the user has to use a dedicated disk for that.
By the way I think periodically flushing logs is not likely to catch much useful errors because the most useful logs are most likely be generated when the system is in extreme condition where flush scripts is not likely to be even ran.