The current jstat gc option (e.g., -gcutil) output columns are too narrow to accommodate large heaps and longer GC times, which makes it difficult to read. Column width changes will have no effect on log parsers, since existing column names will not change.

Sizes are represented in the output in KB to two decimal places. New column widths assuming a 2TB heap should be:

Old: 12 (2/3 of 2TB), was 10
New: 12 (1TB = half of 2TB), was 8
Eden: 12 (1TB = half of 2TB), was 8
Survivor: 11 (256GB = 1/4 of 1TB), was 6
Metaspace: 10 (32GB), was 6
Compressed Class Space: 9 (3GB), was 6
Young GC Count: 6 (1M), was 5 in some cases
Full GC Count: 5 (100K), was 5
Concurrent GC Count: 5 (100K), was 5
Times: 9 (28 hours), was 8 (3 hours)

Sizes should all be right justified rather than sometimes centered, sometimes right justified.

Current jstat tests in test/jdk/sun/tools/jstat use only the default collector, which has been G1 for a long time. When using other collectors, the concurrent cycle column entries are '-' rather than a number. The tests should run the Parallel collector explicitly in addition to the default collector.