ARC1024 SECTOR COUNT may be INCORRECT in JOURNAL DUMP RESTART CASE
During retrieving journal table dump case, the AMPs are sifting through the journal rows putting them into different piles (e.g. A0, B0, etc., type of blocks). The AMP code periodically flushes the blocks to the host but multiple AMPs are doing these flushes at different times. The rowid from each AMP is really only going to reflect the most recent flush but some blocks from different AMPs will be mixed together on the ARCHIVE tape. Therefore, report of how much data archived may be incorrect since restart occurred. There should be no data integrity problems. Its just that the sector/byte counts will be off.
The only way around this is to coordinate between the host and the AMPs as to when the checkpoint is taking and that would require non-trivial changes to both host and AMP and a number of assorted other problems.