Re: [Evms-devel] raid1_mgr plugin patch

[Mike T. <mh...@us...>]

Alexey,

What is in evms-2.1.1-md.patch.bz2 ?  After decomprising the file, I got
evms-2.1.1-md.patch which is 0 size file. Please check.

Mike T.

On Wed, 2003-09-24 at 13:06, Alexey Kotovich wrote:
> Hi Mike;
> 
> I have re-written raid1/5 region mgr again ;)
> 
> There are several places where I have removed calls region re-ordering.
> I can explain why. I think that we should not re-order disks in array
> within evms and let MD perfrom this operation by itself. It means that
> we should keep devices order in array as is. It is important if we want
> to be able to restore degraded regions by adding spare disks even if those
> regions are already mounted. This also will work (I would say it is
> already working) with disk hotswap (scsi/ide).
> 
> Verify and fix array routine had been simplified too. The most important
> thing which we should fix there is device number. The others are not vital
> although I have implemented simple checks there.
> 
> Unfortunately I have not been looking at another useful things like
> raid-reconf, resize and so on due to have lack of time. The main task
> which I wnated to resolve was EVMS stability. It seems I've done it more
> or less.
> 
> BTW, I have some words with regard to DM. Probably it is interested for you.
> I am unhappy with non-persistent device minor number. Sometimes on reboot
> it gets raid resync triggered due to device number replacements. But it is
> really minor issue ;) The major issue which I discovered there is DM_LIST_DEVICES
> ioctl. It constantly returns shrinked devices list. I am wondering how do you
> workaround it in your dm_get_devices_v4? :)
> 
> Anyway, I've made a patch against dm-4.0.1 which resolves this issue.
> The other patches are also attached (raid stuff, some utils and so on).
> Those are for your consideration and I would appreciate any comments.
> 
> Regards,
> Alexey                          ICQ:97715595
> --
> Menu, n.:
> 	A list of dishes which the restaurant has just run out of.	
--