|
On 3 Apr 2000, Stephen Zander wrote:
> >>>>> "Joseph" == Joseph Mack <mack@xxxxxxxxxxx> writes:
> Joseph> Can you check the posible data corruption problem
> Joseph> sometime. Like connect directly from a client box to 2
> Joseph> realservers and write to one mysqld and check if the other
> Joseph> mysqld can see the write
>
> That will work: as soon as both boxes write, however, the database
> will corrupt.
can you explain this a bit more. Something worked (presumably what I
asked, the 2nd box could read the first write) but something did not -
both boxes writing - but what did you do that didn't work? Did you do the
2nd write while the first connection was still open? (we need to know
whether that works).
Presumably you didn't close the connection to the database from the first
client, when you read from the 2nd client (which is reasonable, reads
should be benign). Can you do 2 writes if you close the connection from
the first client, before doing the 2nd write. If so then all writes can be
serialised provided that the connections to the database concerned is
dropped between writes. To get it to work, do you have to close the
connection to mysqld on the first client, or just disconnect from the
particular database?
> mySQL does *not* provide any transaction support of
> concurrency access.
I guess we shouldn't be too surprised here. I doubt if TcX ever expected 2
independant mysqld's to be writting to the same database.
Joe
--
Joseph Mack mack@xxxxxxxxxxx
|
