> On 9 Dec 2015, at 17:35, David Pacheco <[email protected]> wrote:
> 
> On Wed, Dec 9, 2015 at 1:25 AM, George Mamalakis <[email protected]> wrote:
> On 08/12/2015 06:51 μμ, Robert Fisher wrote:
>> IMO nothing should ever depend on boot order of nodes. ​If your application 
>> requires this, its architecture is wrong. If you need A up before B, that 
>> strongly says to me that the application running on B will die horribly 
>> if/when A goes away, even temporarily. I've seen distributed systems built 
>> this way due to devs taking shortcuts and it bit them horribly further down 
>> the line.
>> 
>> Staggering VM startup, to prevent a "thundering herd" is a different, far 
>> more valid, problem.
>> 
>> R
> There are situations, like boxes with ldap'ed nswitch mechanisms, where the 
> LDAP server needs to boot before these boxes in order for them to acquire the 
> appropriate information and be able to start their services properly.
> 
> 
> 
> What happens if the LDAP server fails after that system is already up and 
> running?

You should be replicating to another LDAP server, and teaching the application 
how to fail over between servers. I’m pretty sure the NSS/PAM LDAP modules 
inherited by Illumos can do that, although they are pretty horrid in most other 
regards.

Chris

-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Reply via email to