Re: Communications link failure due to underlying exception...various network exceptions

[mreichman]

Mark Matthews Wrote:
-------------------------------------------------------
> mreichman Wrote:
> --------------------------------------------------
> -----
> > Mark Matthews Wrote:
> >
> --------------------------------------------------
>
> > -----
> > > What happens if you run with
> > > "useServerPrepStmts=false"? It looks like the
> > > connection is going away while the driver is
> > > trying to read results, which often is one of
> > two
> > > things...The server is crashing, or your
> client
> > is
> > > reading rows slower than the timeout
> > > 'net_write_timeout' on the server.
> > >
> > > Disabling server-side prepared statements
> > usually
> > > takes care of the former (and in fact,
> they're
> > > disabled by default in 5.0.5 and later).
> > >
> > > -Mark
> >
> > I'm just starting to experiment with that
> option.
> > If the server crashed, it would be indicated in
> > the log or at least Windows Event Log, right?
> How
> > would I tell about the second possible
> condition?
>
> Yes, you should see server crashes or unexpected
> restarts in either the error log or Windows Event
> log (depending on version and how you have it
> configured).
>
> As for the second possible condition, you'd either
> have to "know" your application, i.e. is it using
> "streaming" result sets, does it run into long
> stop-the-world GC events, etc. or you can change
> the value of net_write_timeout on the server to
> something very high and see if the problem goes
> away.

I'll try that out, thanks. We have a mix of both of those types, so it's worth looking at.
Side note: This machine was on public net and had MySQL exposed. I firewalled it better yesterday, and so far things are smooth. It's hard to tell with one day, but could attempted malicious traffic knock it offline?

>
> >
> > Finally, is there an ETA for when 5.0.5 is out?
>
> Within a week or so, it's just going through
> testing now (there's a matrix of around 25,000
> tests on various platforms, MySQL server versions
> and JDK versions that we run), and once we get
> that all passing we'll do a release. It usually
> takes a week or so to get through this process.
>
> -Mark

That's great news. Besides the hibernate fix mentioned on the changes page and the option default change, are there any other major changes?