[Koha-devel] mysqldump & restore issues

Tue Sep 21 16:11:11 CEST 2010

It could still be memory, because MySQL keeps indexes and such in memory.
It could also be a bad sector on the disk, because MySQL caches query
results to disk when they are bigger than the space allocated in memory for
query results.

The only character I can think of that would cause trouble in the dump would
be an unescaped single quote, or maybe half of a UTF16 character that
happens to have the same value as an ASCII single quote.  It seems unlikely
to me though that mysqldump wouldn't spot such a character as it was
dumping.

Have you tried something like 'select * from biblioitems limit 41537,1'.
That should pull the specific row that has the error, and you could visually
examine it, or see if MySQL throws an error on the row.  That could tell you
if the problem is in the data file specifically, or if it's somewhere else
(memory or disk spaced used to cache the result).

I doubt it's the disk overrunning what it can physically allocate, since as
you say the query results can't be more than a few gigabytes.  Unless the
disk is already nearly full, which I'm sure you've already checked.

I think MySQL could try to allocate memory beyond what it can, beyond the
3.x GB addressable by a 32bit app, but the error message indicates it's a
problem with the disk drive.  Assuming the error message is accurate I'd
start with a bad block scan on the host's hard drive.  That's what I
recommend.

Good luck.

On Tue, Sep 21, 2010 at 7:39 AM, Christopher Curry <ccurry at amphilsoc.org>wrote:

>  Mike,
>
> Thanks for the reply.  I did try querying individual records around that
> row; in fact, I used "limit 41536,1" to find out the specific
> biblioitemnumber.  I checked the record in Koha and found no problems with
> it or adjacent records and I had no problem querying these individual
> records with select statements.
>
> I did try a few queries like "select * from biblioitems limit 1,41537" and
> did get the same 2013 error when querying large datasets, but not when
> querying smaller data sets.  "limit 1,41535" also threw the error, but
> "limit 1,20,000" did not.  I thought this might indicate a memory issue, but
> the syslog error: "InnoDB: Database page corruption on disk or a failed file
> read of page 58164" led me to think there was corrupted data in the
> database.
>
> The VM has 120 GB of dynamically expanding storage and the vdi is housed on
> a filesystem with 108GB free, so there shouldn't be a problem with with
> running out of disk space.  The dump is less than 2GB, so the database
> itself can't be much larger than that, can it?
>
> The VM currently has 1.5 GB allocated for memory, and I tried increasing
> this to 2GB, which did not prevent the error.  The host is 32-bit and is
> maxed out at 4GB of memory, so I can't go much higher than this without
> destabilizing my host.
>
> I had another theory that there might be an offending character in one of
> the MARCXML records that is messing with the format of the SQL commands in
> the .sql file.  Anyone know if this is a possibility?  I'm no SQL expert, so
> I'm not sure what characters to look for.
>
>  Cheers,
>
> Christopher Curry
> Assistant Technical Librarian / Assistant IT Officer
>
> American Philosophical Society
> 105 South Fifth Street
> Philadelphia, PA 19106-3386
> Tel. (215) 599-4299
>
> ccurry at amphilsoc.org
> Main Library number: (215)440-3400
> APS website: http://www.amphilsoc.org
>
>
>
> On 09/20/2010 06:17 PM, Mike Hafen wrote:
>
> Sorry to hear you are having problems with your failover server.  I've had
> problems with failovers servers and backups before, it's not fun.
>
> Have you checked the memory and hard drive of the failover server?  (the
> host)  That would be my first guess.  Also, it may be the filesystem on
> either the host or the guest, but I'm not up on file systems so I can't say
> for sure.
>
> There's a good chance the row number in the error message relates to the
> biblioitemnumber.  You could try just querying the database for rows around
> that.  Something like 'select * from biblioitems where biblioitemnumber >
> 41470 and biblioitemnumber < 41600.  If you see the error it could be data,
> disk, or filesystem related.  If you don't it's probably memory related, but
> maybe filesystem related.
>
> Good luck.
>
> 2010/9/17 Christopher Curry <ccurry at amphilsoc.org>
>
>> Hello all,
>>
>> I'm trying to create a mirrored failover server with data from our live
>> Koha.  In order to do so, I'm using *mysqldump* and *mysql* for backup
>> and restore.
>>
>> I've discovered a troubling problem and I can't determine the cause.
>>
>> I run this command to backup the live server:
>>
>> *mysqldump --single-transaction -ukoha -p koha >
>> /home/koha/KohaServerBackups/koha.`/bin/date +\%Y\%m\%d\%H\%M\%S`.sql*
>>
>> This seems to work correctly (and very quickly!  1.7 GB database exports
>> in 2 min)
>>
>> Then, I run this command:
>>
>> *mysql -v -ukoha -p koha <
>> /home/koha/KohaServerBackups/backupFileName.sql*
>>
>> This also seems to work, as I receive no warnings or error messages.
>>
>> I'm exporting from a 3.00.05.001 system and importing to a 3.00.05.003
>> system, so I then run the *$KOHA_SOURCE/installer/data/mysql/
>> updatedatabase.pl* script.
>>
>> Relevant specs:
>>
>> MySQL version: mysql Ver 14.12 Distrib 5.0.51a, for debian-linux-gnu
>> (i486) using readline 5.2
>> OS: Debian Lenny
>>
>> Failover server is virtual, running on VirtualBox 3.2.8, on an Ubuntu 10.4
>> host.
>>
>>
>> All GUI functions of the Koha failover server seem to operate correctly,
>> but when I run *mysqlcheck -ukoha -p koha* the check fails on
>> koha.biblioitems with the following error message:
>>
>> mysqlcheck: Got error: 2013: Lost connection to MySQL server during query
>> when executing 'CHECK TABLE ... '
>>
>> If I try mysqldump, I get the same error, but it is more specific,
>> reporting that it falls on row 41536.
>>
>> If I check /var/log/syslog, I see this http://pastebin.com/YuuFBHry
>>
>> "InnoDB: Database page corruption on disk or a failed file read of page
>> 58164", etc.
>>
>> Both mysqlcheck & mysqldump work without error on the live server, so I'm
>> thinking that something must be happening to the data during the export or
>> import that corrupts the InnoDB data, but this is speculation, since I'm not
>> a MySQL expert.
>>
>> Has anyone seen behavior like this?  Any suggestions for further
>> troubleshooting/resolution?
>>
>> --
>>
>> Cheers,
>>
>> Christopher Curry
>> Assistant Technical Librarian / Assistant IT Officer
>>
>> American Philosophical Society
>> 105 South Fifth Street
>> Philadelphia, PA 19106-3386
>> Tel. (215) 599-4299
>>
>> ccurry at amphilsoc.org
>> Main Library number: (215)440-3400
>> APS website: http://www.amphilsoc.org
>>
>>
>>
>> _______________________________________________
>> Koha-devel mailing list
>> Koha-devel at lists.koha-community.org
>> http://lists.koha-community.org/cgi-bin/mailman/listinfo/koha-devel
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: </pipermail/koha-devel/attachments/20100921/264aa329/attachment-0001.htm>