[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [ogfs-dev]Thinking about clustering




> -----Original Message-----
> From: Zhu, Yi 
> Sent: Friday, August 01, 2003 5:44 AM
> To: opengfs-devel@lists.sourceforge.net
> Subject: RE: [ogfs-dev]Thinking about clustering
> 
 
> > Not knowing how OpenDLM works ... does only one OpenDLM 
> > node do the lock cleanup?? 
> 
> No, only one node cannot do that. OpenDLM needs all nodes 
> engaged in to get
> the whole picture of the locks -> lock resources map. In OpenDLM
> implementation, the recovery part is the most difficult one 
> (5 sync points
> and 14 recovery steps are used). Basically, when the cluster manager
> notifies OpenDLM there is a node dead, OpenDLM will raise its 
> communication
> priority, at this time all normal lock requests are queued 
> locally and only
> recovery messages are transfered in the cluster -- DLM enters 
> the recovery
> process. During the recovery, the lock resources on the dead 
> node may be
> takeovered by any of the other nodes depending on if they 
> hold the locks of
> that resource and the record on the directory node (something isn't
> important here). After the recovery, communication priority 
> drops down so
> that all queued lock requests are processed.

Thanks for all that good information.  

BTW, is there a document that describes how ODLM works internally (such as the info you've given us here)?  I've seen the API document, which is very useful reading, but IIRC does not describe how the code actually works internally.  Peter B mentioned a doc a few mails back ... was it the API doc, or something else??

> 
> OpenDLM dosen't provide any methods that lock client can 
> interrupt it in
> the recovery process. 

I don't think we would need to interrupt the lock recovery.  We just need to be told when it is complete, and which journal to replay.

> However that does not to say we cannot 
> perform ogfs
> journal replay at the point.

Yes, after lock recovery is complete.

> I think Daniel's "deadman" lock is a very
> smart way to resolve the problem.

Sounds good.  Do you know how to create a "deadman" lock with OpenDLM?

-- Ben --

Opinions are mine, not Intel's

> -----------------------------------------------------------------
> Opinions expressed are those of the author and do not represent
> Intel Corp.
> 
> Zhu Yi (Chuyee)
> Intel China Software Lab (ICSL)
> 22nd Floor, ShanghaiMart Tower No. 2299 Yan'an Road(West)
> Shanghai 200336, PRC
> Tel: 8621-52574545-1261
> Fax: 8621-62366119
> 
> GnuPG v1.0.6 (GNU/Linux)
> http://cn.geocities.com/chewie_chuyee/gpg.txt or
> $ gpg --keyserver wwwkeys.pgp.net --recv-keys 71C34820
> 1024D/71C34820 C939 2B0B FBCE 1D51 109A  55E5 8650 DB90 71C3 4820
> 


-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01
_______________________________________________
Opengfs-devel mailing list
Opengfs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opengfs-devel


[Kernel]     [Security]     [Bugtraq]     [Photo]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Clusters]     [Linux RAID]     [Yosemite Hiking]     [Linux Resources]

Powered by Linux