Hi,

Until this problem is fixed, it might make sense to use the PageStore
storage engine, by appending ";mv_store=false" to the database URL.

Regards,
Thomas


On Friday, May 8, 2015, Mikael Nordenberg <[email protected]> wrote:

> Thank you for your response and time to investigate this.
>
> That is very interesting!
> I attach another case when this happened, in case it can be used to verify
> your assumption.
> I have not been able to reproduce this on a test machine yet.
> But it has happened twice in just a matter of days on one client machine,
> so it may be reproducible on that PC.
> (The affected PC runs Windows 7, and uses a Samsung 850 EVO SSD drive if
> that matters.)
>
> /Mikael
>
>
> Den fredag 8 maj 2015 kl. 18:12:07 UTC+2 skrev Thomas Mueller:
>>
>> Hi,
>>
>> Thanks a lot! I have analyzed the database file, and I think I know what
>> the problem is. It looks like the disk (or operating system) re-ordered
>> write operations, so that changes later in time (and later in the file)
>> were written, but one earlier change (both in time and in the file) was not
>> written.
>>
>> (Detail for later reference: various entries (in chunks 'aad4' to 'aadb',
>> from 16:05:10.378 to 16:05:13.916) think that chunk 'aad3' (later than
>> 16:05:09.369, earlier than 16:05:10.378) is in block 2, but block 2
>> actually contains the earlier chunk 'aa7b' from 16:04:24.742).
>>
>> The MVStore should detect this and automatically discard chunks that were
>> written later than 16:05:10.378. It does not currently do that, this is a
>> problem. I will implement this for the next release.
>>
>> I think what also can happen is write re-ordering causes a file to be
>> truncated too early. This would explain a different corruption problem. The
>> fix for that would need to be different however (truncation would need to
>> wait for 45 seconds).
>>
>> Regards,
>> Thomas
>>
>>
>>
>>
>> On Thursday, May 7, 2015, Mikael Nordenberg <[email protected]> wrote:
>>
>>> Same corruption happened again today on a clients computer when power
>>> disappeared.
>>> Will try to reproduce tomorrow in our lab.
>>>
>>> /Mikael
>>>
>>> Den torsdag 7 maj 2015 kl. 15:42:48 UTC+2 skrev Mikael Nordenberg:
>>>>
>>>> Hi,
>>>>
>>>> We are using H2 version 1.4.185 with mv-store.
>>>> After an unclean shutdown the database file was corrupt. The database
>>>> is attached.
>>>> Opening the database results in http://pastebin.com/PXceifPx
>>>>
>>>> The database is used from a single thread.
>>>> Hope this gives anything...
>>>>
>>>> /Mikael
>>>>
>>>  --
>>> 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]
> <javascript:_e(%7B%7D,'cvml','h2-database%[email protected]');>
> .
> To post to this group, send email to [email protected]
> <javascript:_e(%7B%7D,'cvml','[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