Re: Showstoppers

[Knut Anders Hatlen <Kn...@Sun.COM>]

Dyre.Tjeldvoll@Sun.COM writes:

> Is there an existing bug in Derby which you consider to be a showstopper for
> 10.4? (Meaning that you will veto the release unless this bug gets fixed).
> If this bug is not being actively worked on, please flag it here ASAP,
> so that those looking for bugs to fix can consider it.

After 10.3 was released, we've had a high number of reports about what
seems to be corrupted databases.

In JIRA:

    DERBY-2284
    DERBY-2589
    DERBY-3052
    DERBY-3087
    DERBY-3143
    DERBY-3219
    DERBY-3344
    DERBY-3347
    DERBY-3411

On derby-user:

    http://www.nabble.com/ERROR-XSDG3%3A-Meta-data-for-Container...-----Is-this-database-corrupt--td13697398.html
    http://www.nabble.com/ERROR-XSDG2%3A-Invalid-checksum-on-Page-Page%280%2CContainer%280%2C-1313%29%29-td16389697.html

On derby-dev:

    http://www.nabble.com/Derby-crash-%28urgent%29-td16217446.html

I'm not sure this should block the release, but I can't say I am
completely comfortable with the situation. Does anyone have ideas on how
we can track down these bugs?

-- 
Knut Anders

Re: Showstoppers

[Kathey Marsden <km...@sbcglobal.net>]

Rick Hillegas wrote:
>
> 4) It might be worthwhile for the community to focus on bug fixing for 
> a couple months. That could include a concentrated effort to track 
> down these corruptions.
>
+1

Re: Showstoppers

[Knut Anders Hatlen <Kn...@Sun.COM>]

Rick Hillegas <Ri...@Sun.COM> writes:

Thanks for your input, Rick!

> 3) In general I wouldn't hold up the release of incremental
> improvements because some other issue hasn't been fixed. As far as I
> can tell, 10.4 delivers useful features and performance
> improvements. At first blush, it seems to me to be a better
> distribution than 10.3. I would hold up 10.4 only if we had a good
> reason to believe that 10.4 will make the corruption issues worse.

Agreed. There's nothing indicating that 10.4 is worse than 10.3 in this
respect, so it shouldn't block the release. I just wanted to raise the
issue since bugs that cause database corruption are probably the worst
bugs a database system can have.

> 4) It might be worthwhile for the community to focus on bug fixing for
> a couple months. That could include a concentrated effort to track
> down these corruptions.

+1, sounds like a good idea. In the lack of reproducible test cases I
guess we'll have to focus on code review. If we think the corruptions
are caused by multi-threading bugs, using techniques like adding calls
to sleep() strategic places in the code (see DERBY-3393) could perhaps
also help us smoke them out.

-- 
Knut Anders

Re: Showstoppers

[Rick Hillegas <Ri...@Sun.COM>]

Knut Anders Hatlen wrote:
> Dyre.Tjeldvoll@Sun.COM writes:
>
>   
>> Is there an existing bug in Derby which you consider to be a showstopper for
>> 10.4? (Meaning that you will veto the release unless this bug gets fixed).
>> If this bug is not being actively worked on, please flag it here ASAP,
>> so that those looking for bugs to fix can consider it.
>>     
>
> After 10.3 was released, we've had a high number of reports about what
> seems to be corrupted databases.
>
> In JIRA:
>
>     DERBY-2284
>     DERBY-2589
>     DERBY-3052
>     DERBY-3087
>     DERBY-3143
>     DERBY-3219
>     DERBY-3344
>     DERBY-3347
>     DERBY-3411
>
> On derby-user:
>
>     http://www.nabble.com/ERROR-XSDG3%3A-Meta-data-for-Container...-----Is-this-database-corrupt--td13697398.html
>     http://www.nabble.com/ERROR-XSDG2%3A-Invalid-checksum-on-Page-Page%280%2CContainer%280%2C-1313%29%29-td16389697.html
>
> On derby-dev:
>
>     http://www.nabble.com/Derby-crash-%28urgent%29-td16217446.html
>
> I'm not sure this should block the release, but I can't say I am
> completely comfortable with the situation. Does anyone have ideas on how
> we can track down these bugs?
>
>   
Hi Knut,

I don't have any bright ideas, just some comments:

1) It looks to me as though 2284 and 2589 are 10.2 issues. They are 
marked as 10.3 issues but they seem to have been logged several months 
before 10.3 was released.

2) I agree that these corruptions are disturbing. We may be seeing more 
of them because the engine code is getting creakier. Alternatively, we 
may be seeing more of them simply because Derby usage is ticking up. 
Hard to say.

3) In general I wouldn't hold up the release of incremental improvements 
because some other issue hasn't been fixed. As far as I can tell, 10.4 
delivers useful features and performance improvements. At first blush, 
it seems to me to be a better distribution than 10.3. I would hold up 
10.4 only if we had a good reason to believe that 10.4 will make the 
corruption issues worse.

4) It might be worthwhile for the community to focus on bug fixing for a 
couple months. That could include a concentrated effort to track down 
these corruptions.

Thanks for compiling this list of issues.

Regards,
-Rick