Hi,

By the way, the "corruption after out of memory" is now fixed in the trunk.

Regards,
Thomas


On Tuesday, June 9, 2015, Thomas Mueller <[email protected]>
wrote:

> Hi,
>
> I have now a test case where a MVStore file can not be opened after an out
> of memory. It doesn't look like the file is corrupt, it's just that the
> recovery code can not deal with this case yet. The stack trace is:
>
> Caused by: java.lang.IllegalStateException: Chunk 2 is invalid [1.4.187/6]
>   at org.h2.mvstore.DataUtils.newIllegalStateException(DataUtils.java:773)
>   at org.h2.mvstore.MVStore.readStoreHeader(MVStore.java:695)
>   at org.h2.mvstore.MVStore.<init>(MVStore.java:348)
>
> If you have stack traces of your failures (even better MVStore files),
> that would be great.
>
> Regards,
> Thomas
>
> On Saturday, June 6, 2015, Klaus Wuestefeld <[email protected]
> <javascript:_e(%7B%7D,'cvml','[email protected]');>> wrote:
>
>> Thanks Thomas! I believe simulation is the way to go. It can give you
>> more coverage than trying to perform real power failures.
>>
>> On Fri, Jun 5, 2015 at 1:08 PM, Thomas Mueller
>> <[email protected]> wrote:
>> > Hi,
>> >
>> > The test interrupts the thread, which will close the database file (as
>> > always with FileChannel). Subsequent close calls will fail, this is a
>> bug
>> > that I need to fix. However, I don't think this will corrupt the
>> database
>> > file.
>> >
>> > Interrupting the thread is one way to test it, killing the process is
>> > another. There are already such tests, and so far I couldn't reproduce
>> the
>> > problem.
>> >
>> > Additional tests are with a file system implementation that simulates
>> > re-ordering writes, and power failure. Simulating power failures I
>> have, but
>> > re-ordering writes is still missing.
>> >
>> > Then, real power failure tests are needed. This will require some work,
>> but
>> > I think it is important. I will work on that.
>> >
>> > Regards,
>> > Thomas
>> >
>> >
>> >
>> > On Tuesday, June 2, 2015, Nicolas Fortin (OrbisGIS) <
>> [email protected]>
>> > wrote:
>> >>
>> >> Hi,
>> >>
>> >> I try to write the unit test with thread killing. As I can't reconnect
>> to
>> >> the DB I'm facing some problem to check if the database is corrupted.
>> (see
>> >> attached traceback)
>> >>
>> >> https://github.com/nicolas-f/h2database/tree/thread_kill_mvstore
>> >>
>> >> As I said before, I want to reproduce a problem of corrupted db we
>> have in
>> >> production when using this kind of "brutal" thread kill.
>> >>
>> >> regards,
>> >>
>> >> --
>> >>
>> >> Nicolas Fortin
>> >> IRSTV FR CNRS 2488
>> >> GIS        http://orbisgis.org
>> >> Spatial DB http://h2gis.org
>> >> Noise      http://noisemap.orbisgis.org
>> >>
>> >>
>> >>
>> >> Le jeudi 28 mai 2015 22:04:22 UTC+2, Thomas Mueller a écrit :
>> >>>
>> >>> Hi,
>> >>>
>> >>> That's very interesting.
>> >>>
>> >>> So far I know about a potential problem in case of power failure, and
>> I'm
>> >>> working on fix for that. This is related to write re-ordering.
>> Recovery is
>> >>> affected, and truncating the file needs to be delayed, but both
>> should be
>> >>> quite simple to fix.
>> >>>
>> >>> However, the problem you describe is not related to power failure. I
>> >>> would be very interested in a test case!
>> >>>
>> >>> Regards,
>> >>> Thomas
>> >>>
>> >> --
>> >> You received this message because you are subscribed to the Google
>> Groups
>> >> "H2 Database" group.
>> >> To unsubscribe from this group and stop receiving emails from it, send
>> an
>> >> email to [email protected].
>> >> To post to this group, send email to [email protected].
>> >> Visit this group at http://groups.google.com/group/h2-database.
>> >> For more options, visit https://groups.google.com/d/optout.
>> >
>> > --
>> > You received this message because you are subscribed to a topic in the
>> > Google Groups "H2 Database" group.
>> > To unsubscribe from this topic, visit
>> > https://groups.google.com/d/topic/h2-database/FPyYtNPbgJQ/unsubscribe.
>> > To unsubscribe from this group and all its topics, send an email to
>> > [email protected].
>> > To post to this group, send email to [email protected].
>> > Visit this group at http://groups.google.com/group/h2-database.
>> > For more options, visit https://groups.google.com/d/optout.
>>
>>
>>
>> --
>> Valeu, Klaus.
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "H2 Database" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected].
>> To post to this group, send email to [email protected].
>> Visit this group at http://groups.google.com/group/h2-database.
>> For more options, visit https://groups.google.com/d/optout.
>>
>

-- 
You received this message because you are subscribed to the Google Groups "H2 
Database" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/h2-database.
For more options, visit https://groups.google.com/d/optout.

Reply via email to