10.9.1 Performing compaction on cache data files

(a) When you delete data in the order in which the data was stored

When you delete (remove) data in the order in which the data was stored (put), invalid data is created in cache data files each time data is deleted. In such a case, perform compaction on the cache data files whose entire contents are invalid data. Execute the eztool compaction command with the --threshold option specifying a threshold of 100%.

Reference note

If the order in which the data was stored (put) does not exactly match the order in which data is deleted (remove), set the threshold to a value that is smaller than 100%.

(b) When you store or delete data randomly without regard to the order in which the data was stored

When you store (put) or delete (remove) data randomly without regard to the order in which data is stored (put), the file usage rate will differ from one cache data file to another. For this reason, you need to allocate unused files so that data can be added and deleted. In this case, perform compaction periodically based on the number of unused files. Execute the eztool compaction command with the --unused_fc option specifying the number of unused files to be allocated.

If the planned number of unused files cannot be obtained after compaction processing, reduce the threshold with the --threshold option and then perform compaction again.

(c) Allocating space to cache data files immediately

When the number of unused files is very small and you need to allocate space to the cache data files immediately, perform compaction with a smaller threshold specified in the --threshold option. Execute the eztool compaction command with the priority for caches and ranges specified in the --cache and --range options.

If compaction is already underway, stop the compaction processing and then re-execute the eztool compaction command. For details about how to stop compaction processing, see 10.9.2 Stopping compaction processing.

Reference note

If the amount of available space in the cache data files is very small, we recommend that you close the cluster before performing compaction processing, if possible.