>>> Tim Serong <[email protected]> schrieb am 11.07.2011 um 15:51 in Nachricht
<[email protected]>:
[...]
> You probably want to flip the colocation constraint:
> 
>    colocation col_rksapr00_saprouter_ping inf: \
>      prm_rksapr00_ping grp_rksapr00

Yes I did that once I realized that it's directed.

[...]
> Have a look at 
> http://www.clusterlabs.org/doc/en-US/Pacemaker/1.1/html/Pacemaker_Explained/s-
>  
> resource-colocation.html 
> and the following few pages.  Also Colocation Explained at 
> http://clusterlabs.org/wiki/Documentation.
> 
> The key point being "you can't place A relative to B unless you know 
> where B is", so colocation with Pacemaker necessarily involves a 
> dependency.  AIUI implementing fully "symmetric colocation" would mean 

Note: The "colocation explained document" is a void link in there!

I have nothing against a dependency, but something against a "running 
dependency":

Treated symmetrically, it does not matter who is running first, A or B. Only 
when the other (A or B) comes up, the colocation should be considered, either 
placing the newer resource closer or further away from the other resource.

> Pacemaker may need a semi-infinite amount of time to figure out WTF a 
> given colocation constraint means.  This is undesirable.

Yes, it's like placing a block optimally in a filesystem. Fortunately most 
filesystems need little time and don't place the blocks that badly.

Does the current system implement transitivity, i.e. colocation(A,B), 
colocation (B,C) -> colocation(A,C). I guess that's one root of the problems. 
Also when requiring disjunct sets of colocations the problems should be easy to 
handle.

When ignoring the running state, "colocation foo A inf: B" would be fulfilled 
if B is not up. When B comes up, it would have to run where A is, or A should 
be migrated to where B started. The +/- infinity cases should be easy to handle 
at least.

I'm just wondering if a "first match" rule for colocations would help us here 
(Still there are other constraints like priority and location which make things 
more complicated)

Regards,
Ulrich


_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to