[MarkLogic Dev General] About transaction log and recovery
wayne.feick at marklogic.com
Tue Jun 22 23:28:33 PDT 2010
I'm not sure if there are any docs available outside the company.
Generally speaking, it's what you'd expect in a transactional system;
recovery records are written to the journal as a transaction progresses,
and important records (prepare, commit) cause journal data to be synced
to disk and wait for the sync to complete before proceeding.
On 06/22/2010 11:15 PM, Ling Ling wrote:
> Hello Wayne,
> Thank you for your reply. Yeah, I found it. They are together with data,
> on our striped volume. I am interested in how they are managed, like
> when to write log and when the I/O happens. Could you point out some
> materials that I can read about? Thanks!
> Wayne Feick wrote:
>> Hi Ling,
>> Yes, we maintain a transaction journal for each forest that allows us to
>> recover committed transactions in the event of a failure. In each
>> forest's Journals directory you'll files named Journal# where # is an
>> integer. They are not human readable.
>> On 06/22/2010 10:24 PM, Ling Ling wrote:
>>> When I look at the logs, I only found OS logs and ML server file logs.
>>> These log told what ML server did. In traditional database, there are
>>> redo/undo logs when database writes. Does ML server writes such logs?
>>> Where is it and when does the server write such logs and flush it to
>>> disk? Is it related to the version control in ML server? We are doing
>>> some update and insert with ML server and want to make sure what happens
>>> here. Thank you very much.
>>> General mailing list
>>> General at developer.marklogic.com
> General mailing list
> General at developer.marklogic.com
Phone +1 650 655 2378
Cell +1 408 981 4576
More information about the General