Re: Failed consistency check

Thanks for those of you who replied. We are using OnLine 5.02UC3 on SCO 3.2V4.0. In our case, the problem is not in the shared memory, and it is not caused by chunk overlapping either. It is caused by some leftover on the hard disk. The tables involved are usually small tables -- only 3 pages are used: one bitmap page, one index page, one data page, and the data page is not full. ( we use default extent size, so 8 pages are allocated for this table). Somehow on the 4th page, the 2 bytes page flags within the 24 bytes page header ended with "0x9", which cause OnLine to think this is a "remainder page", which is not true for this table.(for this table, the 4th page should have page type "unknown", because it is not used.) That is why "tbcheck -pt" gives me that error message. I find a quick way to fix this, but it involves some binary work. I calculated the physical address of this page on the hard disk, modified the page flags byte on the hard disk so it does not end with "0x9"(whatever makes OnLine think this page is "unknown" type), stopped and restarted OnLine, and the error went away. But I am sure this method is extreamly dangerous and highly unrecommended. In our case the error hasn't caused any real harm other than sending the annoying message to online.log. I am thinking (haven't tried) that by inserting more rows into this table(so OnLine will overwrite the 4th page), then delete unwanted rows might also work. I am waiting for the error to reappear to try this out. (I hope I don't need to, though). Thanks, Zhenling