Sometimes Can't Kill Running Query

[vygi]

Hello here,

I am not absolutely sure if it always happens (can't verify right now) but I am pretty sure that there is following problem:

– start several connections to several SQL servers.

There must be several different databases on them, eg. DB1_1 on the 1st and DB2_1 and DB2_2 on second server.

– start some slow query on 1st connection which takes longer that 5-10s to execute;

– now switch to the 2nd connection and change here to some another database which does not exist on the 1st server, eg. to DB2_2

– switch back to the 1st connection and try to kill running query. SQLyog will say “database DB2_2 does not exist”.

Surely it doesn't! It is on the 2nd server!

Obviously SQLyog does not tell first connection from second in this sase.

Hope it helps,

Vygi

[peterlaursen]

Please tell how you connect!

I briefly tried to reproduce this but I cannot.  

We will now try more systematically!

[peterlaursen]

This is reproduced (and partially explained) now.

The problem is that when you click a database in second connection a USE statement is sent.

If you swith back to first connection by clicking the title bar of the connection window (and doing nothing else before trying to kill the connection) no new USE statement is sent. The USE statement from second connection is still the active one.

But USE statements should be specific for each connection, I think.

And personally I am not sure why we need to address some specific database here at all!

We will study our own code again! 

[vygi]

well… it sounds really weired … because two different connections are mixed up:

this USE statement should be run on another connection and do not affect the first one.

Usually it is speciffic to connection but seems like it does not work like expected when a slow query is running. And everyithing is again OK when it's finished.

Anyway, I second that you should study your own code, as I can't explain what is going wrong here.

[peterlaursen]

Bikas is debugging his way through the code right now.

With all operations on data the program can have more USE statements 'in the air' simutaneoulsy and use the correct one.

What exactly happens with this STOP issue I cannot tell now, but probably can tomorrow.

This issue (and two very small cosmetical issues) are all what we are missing before 5.23 RC1 can be released. So it is priority now actually!

[peterlaursen]

Now fixed in the development tree.

RC1 with the fix is expected in 1-2 days.

[Author]

Posts