Subject Re: [firebird-support] Re: FB deadlock during big table indexing on Linux
Author Bogusław Brandys
taliyev wrote:
> --- In firebird-support@yahoogroups.com, Helen Borrie <helebor@...>
> wrote:
>>
>>>> At 02:56 AM 23/04/2006, you wrote:
>>>>> We have huge database ( 30 GB) that seems to be working ok on
>>>>> Windows. We need to port this database to Linux. Transportable
>>>>> backup file is about 20GB in size. Biggest table is about 185
>>>>> millions of records.
>>
>>
>> At 11:56 AM 23/04/2006, you wrote:
>>> Our Linux is CentOS 4.3 (RedHat Enterprise clone) with ext3
>>> partition. We have configured /tmp to be used for temp files and
>>> this directory spans across entire disk space. We have about 90 GB
>>> available for it to grow.. FB has `rwx' permissions there…
>>>
>>> We have also tried to increase SortMemBlockSize and
>>> SortMemUpperLimit with no success. Any other suggestions?
>> Not a lot. What data type is the primary key defined as in your
> metadata?
>> PLEASE DO NOT TOP POST WHEN YOU REPLY TO POSTINGS !!
>> Edit the existing text and place your reply within the text and/or
> at the end.
>> ./heLen
>>
>
> This particular table does not have PK defined. Indexes we have are
> on VARCHAR(3 or 4) fields and are non-unique, but high selectivity.
> It seems that FB looses it's handle on temp files… I do not know
> why…. Somehow Windows FB version can handle those indexes and Linux
> cannot...
>

Maybe try to make other directory for temp space and use that for
testing.If you would restore using that other temp directory then
probably something is messing with /tmp directory (cron job?)


Regards
Boguslaw