There are various occasions where we’ve encountered disk space issues such as HELP-38478. To aid the diagnosis of these issues it may be useful to extend some of the statistics with the block manager. This will also help understand the effectiveness of future changes to the block manager such as WT-10028 and help in tests such as WT-10014. We can use this ticket as a placeholder for refinement work and future ideas along these lines. Some extensions to consider:

1. Count the number of blocks we skip when looking for a block that ‘best fits’. This could be compared with the number of blocks read to see if we’re skipping many blocks for few block searches.
2. Expose some of the block manager statistics to the connection level to view in T2, for example we could view the rate of a statistic like block_extension which counts the number of blocks that require a file extension.
3. Add a statistic at both the datasource and connection level to count the number of blocks reused and/or the bytes that a reused.
4. Add a statistic block_reused_entries to count the number of entries in the available list. This could be used with the block_reuse_bytes to calculate the average size of an available block.
5. The wt utility has a dump_layout function which outputs the extents in buckets such as

extents by bucket: \{4KB: 1}, \{8KB: 2}, \{16KB: 1}, \{128KB: 1}, \{512KB: 3}, \{1MB: 5}, \{2MB: 9}, \{4MB: 13}, \{8MB: 17}, \{16MB: 11}, \{32MB: 6}, \{64MB: 2}

 Could we expose this as a statistic in some way?