Re: Derby on NAS = corruption ?

["Dag H. Wanvik" <da...@oracle.com>]

Mo Maison <mo...@yahoo.fr> writes:

> Hello,
>
> I have read the referenced thread with great interest,
> and I have a related question.
>
> If we replace NAS with SAN (SAN is basically a remote
> block device).
> If a single server mounts the SAN partition, and a single
> derby server is run on this server, I guess we shouldn't
> encounter any trouble, should we ?

Derby uses a write-ahead log to guarantee that committed transactions
are persisted. If the log device is on the network storage, you need a
guarantee that the log write operations are actually
persisted. Similarly, for safe operation one needs to disable disk
cache in the operating system for the log device.

Cf. http://db.apache.org/derby/docs/10.9/adminguide/cadmindbintegrity.html

Dag



>
> Regards,
>
>   M. Maison
>
>
> On 18/07/2012 14:48, Rick Hillegas wrote:
>> On 7/18/12 1:56 AM, Santiago Miguel wrote:
>>> Hi,
>>>
>>> I work in a french company which provides secure online file
>>> management services (SaaS solutions).
>>>
>>> In one of our applications (backup software), in case of a large
>>> file, we cut it into blocks and we use a Derby db in order to
>>> maintain the block indexes. These Derby dbs (version 10.5.1.1) are
>>> stored on a NAS (cluster Isilon NL series) shared between
>>> webservers. The derby db is directly  acceded on NAS and it is
>>> thread safe.
>>> Everyday, we are observing many Derby db corruptions (Unknown page
>>> format at page Page(524,Container(0, 1024))) and these dbs are no
>>> longer usable.
>>>
>>> I join the derby log file corresponding to the last night on 1
>>> webserver.
>>> Thanks in advance for your help.
>>>
>>> Best regards,
>>> Miguel Santiago
>>>
>> Hi Miguel,
>>
>> The following email thread may shed light on the problems you are
>> seeing:
>> http://old.nabble.com/Client-server-plus-network-storage-of-database-files-td19565143.html
>>
>> Hope this helps,
>> -Rick
>>