>>> Dejan Muhamedagic <[email protected]> schrieb am 05.07.2011 um 17:16 in
Nachricht <20110705151610.GA3822@squib>:
[...]
> > Now the funny thing is that you cahnged the DTD in an incompatible way
> without changing the version number. That's very bad practice!
>
> You're right, the version should've been changed. But, to the
> best of my knowledge, the dtd at www.opencf.org has never been
> used.
After searching for the fil in github without success, I found an easier way:
# rpm -ql resource-agents |grep dtd
/usr/share/resource-agents/ra-api-1.dtd
And for the records, here are the diffs between the old and the new version:
--- RA-API/ra-api-1.dtd 2011-07-06 08:38:33.000000000 +0200
+++ /usr/share/resource-agents/ra-api-1.dtd 2011-03-10 15:41:33.000000000
+0100
@@ -1,6 +1,6 @@
<?xml version="1.0" encoding="UTF-8"?>
-<!ELEMENT resource-agent (version,parameters,actions,special) >
+<!ELEMENT resource-agent (version,longdesc,shortdesc,parameters?,actions) >
<!ATTLIST resource-agent
name CDATA #REQUIRED
version CDATA #IMPLIED>
@@ -14,6 +14,7 @@
<!ELEMENT parameter (longdesc+,shortdesc+,content)>
<!ATTLIST parameter
name CDATA #REQUIRED
+ required (1|0) "0"
unique (1|0) "0">
<!ELEMENT longdesc ANY>
@@ -26,18 +27,14 @@
<!ELEMENT content EMPTY>
<!ATTLIST content
- type (string|integer|boolean) #REQUIRED
+ type (string|integer|boolean|time) #REQUIRED
default CDATA #IMPLIED>
<!ELEMENT action EMPTY>
<!ATTLIST action
- name (start|stop|recover|status|reload|meta-data|verify-all)
#REQUIRED
+ name
(start|stop|recover|monitor|restart|migrate_to|migrate_from|promote|demote|notify|status|reload|meta-data|usage|methods|validate-all)
#REQUIRED
timeout
CDATA #REQUIRED
interval
CDATA #IMPLIED
start-delay
CDATA #IMPLIED
+ role CDATA #IMPLIED
depth
CDATA #IMPLIED>
-
-<!ELEMENT special ANY>
-<!ATTLIST special
- tag CDATA #REQUIRED>
-
Still the DTD lacks comments to describe the semantics of the elements. For
inspiration, you might look at the XHTML DTD (just one example):
<!-- attributes for elements that can get the focus
accesskey accessibility key character
tabindex position in tabbing order
onfocus the element got the focus
onblur the element lost the focus
-->
<!ENTITY % focus
"accesskey %Character; #IMPLIED
tabindex %Number; #IMPLIED
onfocus %Script; #IMPLIED
onblur %Script; #IMPLIED"
>
While we are at it: The current specification seems to require "promote" and
"demote":
($Id: ocf-tester,v 1.2 2006/08/14 09:38:20 andrew Exp $)
Beginning tests for ./whatever...
WARN: Can't check agent's permissions because we're not root; they should be 755
* rc=7: Monitoring an active resource should return 0
* rc=2: The notify action cannot fail and must return 0
* rc=2: Demoting a start resource should not fail
* rc=2: Promote failed
* rc=2: Demote failed
whatever[13899]: ERROR: ./saprouter demote not implemented
Aborting tests
The "Dummy RA" however doesn't provide those. Also the docs (dev-guides)
(section 5.6) says the "promote" is optional. Also, I don't see why a non-root
user cannot check the numeric file permissions of the RA. While wondering, I
found the test being implemented as "su nobody -s /bin/sh $agent $action >
/dev/null". In SLES there is a command "stat -c %a file" to print the octal
file permissions. Why not use that? Requiring to be root, just to switch to a
non-root user is somewhat strange.
I just found out that the RA dispatcher must return 3 (not 2) for unimplemented
methods. Of course it would be better if the ocf-tester would actually consider
the meta-data XML actions before trying to start them.
For fun I checked the RAs provided with the system (SLES11 SP1), and I found
one bad RA:
/usr/lib/ocf/resource.d/heartbeat/SAPInstance
-:139: element resource-agent: validity error : Element resource-agent content
does not follow the DTD, expecting (version , longdesc , shortdesc ,
parameters? , actions), got (version shortdesc longdesc parameters actions )
</resource-agent>
^
Any advice on writing a correct RA? (Actually I managed to make my RA pass the
ocf-tester in the meantime)
Reagrds,
Ulrich
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
