Bug 429525 - rgmanager rgmanager detects false collisions between attributes of a vm resource
rgmanager rgmanager detects false collisions between attributes of a vm resource
Status: CLOSED NOTABUG
Product: Red Hat Enterprise Linux 5
Classification: Red Hat
Component: rgmanager (Show other bugs)
5.1
All Linux
low Severity low
: rc
: ---
Assigned To: Lon Hohberger
:
Depends On:
Blocks:
  Show dependency treegraph
 
Reported: 2008-01-21 06:55 EST by Pieter Krul
Modified: 2009-04-16 18:56 EDT (History)
1 user (show)

See Also:
Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Environment:
Last Closed: 2008-01-21 11:07:52 EST
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---


Attachments (Terms of Use)
cluster.conf example (2.37 KB, text/plain)
2008-01-21 06:55 EST, Pieter Krul
no flags Details
Example of how to allow multiple instances of the same domain name (1020 bytes, patch)
2008-01-21 11:05 EST, Lon Hohberger
no flags Details | Diff

  None (edit)
Description Pieter Krul 2008-01-21 06:55:18 EST
Description of problem:
When having multiple <vm> resources with different failover domain, but
different name attributes, rgmanager reports an error "Error storing vm: Duplicate"


Version-Release number of selected component (if applicable):
rgmanager-2.0.31-1.el5

How reproducible:
Always

Steps to Reproduce:
1. Have two vm resources, with same domain, but different names.
2. Update cluster.conf throughout cluster with ccs_tool and cman_tool
3. Examine /var/log/messages
  
Actual results:
clurgmgrd reports an error

Dec 28 11:46:20 dom0 ccsd[5920]: Update of cluster.conf complete (version
2007122127 -> 2007122128).
Dec 28 11:46:25 dom0 clurgmgrd[9762]: <notice> Reconfiguring
Dec 28 11:46:25 dom0 clurgmgrd[9762]: <err> Error storing vm: Duplicate
Dec 28 11:46:26 dom0 clurgmgrd[9762]: <err> Error storing vm: Duplicate

Expected results:
rgmanager should configure this resource without errors

Additional info:
The VM's are started properly. See attached cluster.conf
Comment 1 Pieter Krul 2008-01-21 06:55:18 EST
Created attachment 292354 [details]
cluster.conf example
Comment 2 Lon Hohberger 2008-01-21 10:58:28 EST
Actually, you're looking in the wrong place (and it's my fault for telling you
to look there).  I'm sorry for pointing you in the wrong direction - the errors
aren't actually from cluster.conf - they're from a dirty /usr/share/cluster.

Clean up /usr/share/cluster ...

# cd /usr/share/cluster
# rm -f *~ *.rpm*

rg_test works on that conf file (note that both VMs are displayed below).  I
left some errors (showing duplicate resource agents):

[root@ayanami daemons]# ./rg_test ../resources test ~lhh/futon.conf
Using ../resources as resource agent path
Running in test mode.

Warning: Ignoring ../resources/ip.sh.rpmsave: Bad extension .rpmsave
Error storing fs: Duplicate
Error storing clusterfs: Duplicate
Error storing fs: Duplicate
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ duplicate resource agent errors

Loaded 21 resource rules
=== Resources List ===
Resource type: vm [INLINE]
Instances: 1/1
Agent: vm.sh
Attributes:
  name = vm-xennfs1-pru [ primary ]
  domain = xen_pru [ reconfig ]
  autostart = 0 [ reconfig ]
  hardrecovery = 0 [ reconfig ]
  exclusive = 0 [ reconfig ]
  recovery = disable [ reconfig ]
  path = /guests/virtdata/scripts:/etc/xen
  migrate = live
  depend_mode = hard
  max_restarts = 0 [ reconfig ]
  restart_expire_time = 0 [ reconfig ]

Resource type: vm [INLINE]
Instances: 1/1
Agent: vm.sh
Attributes:
  name = vm-xennfs2-pru [ primary ]
  domain = xen_pru [ reconfig ]
  autostart = 0 [ reconfig ]
  hardrecovery = 0 [ reconfig ]
  exclusive = 0 [ reconfig ]
  recovery = disable [ reconfig ]
  path = /guests/virtdata/scripts:/etc/xen
  migrate = live
  depend_mode = hard
  max_restarts = 0 [ reconfig ]
  restart_expire_time = 0 [ reconfig ]

=== Resource Tree ===
vm {
  name = "vm-xennfs1-pru";
  domain = "xen_pru";
  autostart = "0";
  hardrecovery = "0";
  exclusive = "0";
  recovery = "disable";
  path = "/guests/virtdata/scripts:/etc/xen";
  migrate = "live";
  depend_mode = "hard";
  max_restarts = "0";
  restart_expire_time = "0";
}
vm {
  name = "vm-xennfs2-pru";
  domain = "xen_pru";
  autostart = "0";
  hardrecovery = "0";
  exclusive = "0";
  recovery = "disable";
  path = "/guests/virtdata/scripts:/etc/xen";
  migrate = "live";
  depend_mode = "hard";
  max_restarts = "0";
  restart_expire_time = "0";
}
=== Failover Domains ===
Failover domain: xen_pru
Flags: Ordered Restricted 
  Node node1 (id 1, priority 1)
  Node node2 (id 2, priority 2)
=== Event Triggers ===
Event Priority Level 100:
  Name: Default
    (Any event)
    File: /usr/share/cluster/default_event_script.sl

If it was a question of unique/non-unique, the error would look like this:

Error: Primary attribute collision. type=vm attr=name value=vm-xennfs1-pru
Error storing vm resource

You, however, are correct - rgmanager does not allow configuration of the same
VM name even if restricted failover domains do not overlap.  For example, the
following configuration would result in the above error collision error:

        <rm log_facility="local4" log_level="5">
                <resources/>
                <failoverdomains>
                        <failoverdomain name="domain1" ordered="1" restricted="1">
                                <failoverdomainnode name="node1" priority="1"/>
                                <failoverdomainnode name="node2" priority="2"/>
                        </failoverdomain>
                        <failoverdomain name="domain2" ordered="1" restricted="1">
                                <failoverdomainnode name="node3" priority="1"/>
                                <failoverdomainnode name="node4" priority="2"/>
                        </failoverdomain>
                </failoverdomains>
                <vm autostart="0" domain="domain1" exclusive="0"
name="vm-xennfs1-pru" path="/guests/virtdata/scripts:/etc/xen" recovery="disable"/>
                <vm autostart="0" domain="domain2" exclusive="0"
name="vm-xennfs1-pru" path="/guests/virtdata/scripts:/etc/xen" recovery="disable"/>
        </rm>

This is a known limitation.  It is kind of a design flaw - basically, resource
placement policies and individual resource configuration are not the same thing
in rgmanager.  Fortunately, it's fairly easy to make rgmanager allow multiple
domains with the same name.

Comment 3 Lon Hohberger 2008-01-21 11:05:20 EST
Created attachment 292381 [details]
Example of how to allow multiple instances of the same domain name

Untested.  The idea is that you can provide a non-unique name_override in
cluster.conf which is the -real- Xen domain name for the VM.

Note You need to log in before you can comment on or make changes to this bug.