Re: Avoiding "Repair with keycache"

[Eric Walker]

Ingo Strüwing wrote:

> The theoretical maximum index size for normal
> indexes is maximum field length (times character
> width where appropriate, 3 for UTF8) plus 1 for
> max lengths below 256, 2 otherwise, plus the
> myisam_data_pointer_size, times the current number
> of records in the table. For full text indexes the
> maximum word length (31) (times character width)
> is added to the key length.
>
> I'm not sure if setting the variables to 20G is
> possible and will work. Just give it a try. . . .

It is, I did, and it worked a treat. What had taken over 47 hours before now ran in 2:15 (which means that, as a generalization, the keycache method takes roughly 20 times as long as the sort method, nearly turning each hour of indexing into a day of indexing).

> Regarding the fulltext index size I can neither
> confirm nor reject it. I have not seen much
> fulltext indexes yet. Let alone looked at their
> size.

Thank you again for the info and advice.

Cordially,
Eric Walker
Owlcroft House