On Fri, Nov 07, 2014 at 01:59:37PM -0500, Prarit Bhargava wrote:
On 11/07/2014 12:48 PM, Vivek Goyal wrote:
> On Fri, Nov 07, 2014 at 07:54:12PM +0800, WANG Chao wrote:
>
> [..]
>> If the dump target is dedicated ssh/nfs server and it's configured to be
>> the dump target for many (say hundreds of) systems, there will be a lot
>> of "latest" directories with different host-ip prefix there. I think
>> that's kind of annoying. So I'll go with Vivek.
>>
>> I'd like to keep this feature simple and clear. And that is just
>> creating "latest" for local dumps. How about naming the latest -
>> "latest-localhost" or something like that? Thoughts?
>>
>
> I prefer not to do anything actually. I think it adds more code but does
> not buy us much.
>
> - First of all it works only for local systems. So differentiating between
> two cases can be confusing for users.
>
> - Secondly, I am wondering what will happen in case of clustered lvm
> kind of configuration or distributed file systems like gfs.
>
> - It is really easy to either read time stamps or do "ll -rt" to figure
> out which is the lastest file.
>
> - If somebody is already using a soft link "latest" in /var/crash/ dir
for
> whatever reason, it will conflict with that too.
>
Okay, is there a "hook" in which I can do this automatically in kdump?
Yes. kdump allows you to write your own script which can run after saving
dump. Look at "kdump_post" option.
You can create a softlink there to customize it for your usage.
Thanks
Vivek