Checkpointing your filesystem

System Administration Guide
Chapter 2, Administering filesystems

Checkpointing is the process of transitioning filesystems to a clean (consistent) state. Filesystem data consists of user file data (the contents of a file) and the data structures used to store the data (also known as ``meta data''). Recently-accessed data is held in memory (``cached'') for a short time in case it is needed again. If the system is stopped unexpectedly, this cached data can be lost.

By default, HTFS, EAFS, AFS, and S51K filesystems periodically ``checkpoint'' (write) cached meta data back to disk. This increases the probability that the filesystem meta data will be in a consistent state if the system is halted unexpectedly. (There may be a small loss of user data, which is not checkpointed.)

If your system should experience a system error, checkpointing will reduce the likelihood that the filesystem will need to be checked and repaired with fsck(ADM) when rebooting, thus mimimizing downtime.

NOTE: While checkpointing is appropriate for most users of HTFS, there is a small performance penalty. To achieve maximum throughput, you should consider disabling checkpointing.