[ClusterLabs] Antw: [EXT] How to stop removed resources when replacing cib.xml via cibadmin or crm_shadow

Ulrich Windl Ulrich.Windl at rz.uni-regensburg.de
Thu Oct 1 03:16:23 EDT 2020


>>> Igor Tverdovskiy <igor.tverdovskiy.pe at gmail.com> schrieb am 30.09.2020 um 23:49
in Nachricht
<CAARZOoj15B37dsv+Bfx-8MMNOr4Sh6sT6SPBCeK4ex5Babau-g at mail.gmail.com>:
> Hi All,
> 
> I have a necessity to apply the whole cib.xml instead of using command line
> tools like
>> crm configure ...
> 
> I have generated proper XML according to predefined templates and apply it
> via
>> cibadmin --replace --xml-file cib.xml

I think before doing such a thing you should put all nodes into standby, or stop all resources.
In the past it had been a bad idea to delete running resources or trying to rename such.
When replacing the CIB you probably do such things.
What are your specific reasons to replace the CIB?

> 
> However I have encountered an issue when resources (in particular VIP
> addresses) which were completely removed/replaced by others continue to run
> on the interface after XML replacing.
> 
> Orphaned resources appear and VIP actually is kept started.
> ```
> haproxy-10.0.0.111     (ocf::haproxy):         ORPHANED FAILED 738741-ip2
> (blocked)
> vip-10.0.0.111 (ocf::IPaddr2):         ORPHANED FAILED 738741-ip2 (blocked)
> 
> ```
> 
> What I have tried:
> 1. Replace pure cib.xml without <status> element (only <configuration>)
> 2. Take active CIB from "cibadmin -Q" and replace only <configuration>
> element while keeping <status> as is.
> 3. Create shadow replace cib.xml with/without status and commit.
> Indeed crm_simulate -LS shows intention to stop vip-1.1.1.1, but in fact it
> will not after shadow commit.
> 
> Sometimes I can manage to automatically clear removed/replaced VIP
> addresses from the interface after replacing XML, so it is
> definitely possible, but I can not understand how to achieve this.
> 
> So I wonder is there a way to replace XML but at the same time stop
> resources which are removed in a new XML?
> 
> Thanks!






More information about the Users mailing list