Re: Random data loss, less than 0.1% of the time - how to prevent it

[Ted Byers]

Peter Brawley Wrote:
-------------------------------------------------------
> > If MySQL does store an error log, please tell me
> where to look for it on Windows
>
> <machine_name>.err inn the mysql data dir.
>
OK, I will look tomorrow. I was quite ill today.

> > I have demonstrated that
> Exception::Class::TryCatch is reliable
>
> No, your error thingie passed just one limited
> test.
>
That was only the latest test of many. When I decided to use Exception::Class::TryCatch, I did a suite of tests, checking every kind of error I could imagine: too many to report on here.

> > Why would this fail, even if a cgi script is
> executed for two requests that happen (nearly)
> simultaenously?
>
> It's a long list of low-probability possibilities
> (Rick's deadlock example being one of them).

Is this long list of possibilities documented anywhere (concisely, so I don't have to wade through thousands of pages of text in order to know how to proceed)? It would be good to have such a document to guide creating a series of tests.

> The fact that it is failing occasionally indicates
> that error control needs to be beefed up and maybe
> such writes need to be put in transaction blocks
> with rollback logic.
>
Do you have, or can you point me to, a document on the web that describes such a beefed up error control, especially one to be used when a given set of SQL statements can not be permitted to fail? If you have an insert statement that can not be permitted to fail under any circumstances, and an error happens, how can you handle the error condition and retry the insert so it can't fail?

I would not hesitate to beef up my error handling if I knew what more could be done.

> Let us know if you find anything in the error
> logs?

I will. I just hope I am recovered enough to be able to progress tomorrow.

Thanks

Ted