Bug 1384484
| Summary: | Some resource agents' metadata do not conform to the xml schema | ||
|---|---|---|---|
| Product: | Red Hat Enterprise Linux 8 | Reporter: | Tomas Jelinek <tojeline> |
| Component: | pacemaker | Assignee: | Ken Gaillot <kgaillot> |
| Status: | CLOSED WONTFIX | QA Contact: | cluster-qe <cluster-qe> |
| Severity: | low | Docs Contact: | |
| Priority: | low | ||
| Version: | 8.0 | CC: | cluster-maint, mgrac, oalbrigt |
| Target Milestone: | pre-dev-freeze | ||
| Target Release: | --- | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | If docs needed, set a value | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2020-10-13 21:56:43 UTC | Type: | Enhancement |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
This is a reflection of the state of the OCF standard. The standard was created in 2002 and the group that originally maintained it disbanded, so software packages (including pacemaker) have added custom extensions to it over the years. The proper solution to this issue would be for pacemaker to document its extended schema, and make the ocf:pacemaker resource agents claim adherence to that. (I wouldn't be surprised if there are also a few errors that need to be corrected, but that would the most important thing.) There is some upstream discussion about adopting the OCF standard under the banner of ClusterLabs, and updating it. That would help this issue, but pacemaker would likely still need extensions as experimental features were trialed before being adopted in the standard, so a documented extension schema could still be worthwhile. Also, fence agents are not covered by the OCF resource agent standard. They follow a very similar format, but at least the ones I've looked at don't claim adherence to the DTD, and none meet the standard (fence agents support different actions than resource agents, among other differences). The effort to update the OCF standard will likely incorporate fence agents. @Ken: Can you take a look at our schema: https://github.com/ClusterLabs/fence-agents/blob/master/fence/agents/lib/metadata.rng or example at: https://github.com/ClusterLabs/fence-agents/blob/master/tests/data/metadata/fence_apc.xml - I'm mostly interested on pieces that you are missing. I'm aware that there is a ton of tags that are not important for resource agents as they support cmdline interface as much as fence agents. Fence agents metadata obtained from pacemaker do not conform to the schema shipped with fence agents due to extra actions added by pacemaker in the stonith_api_device_metadata function in the st_client.c file:
[root@rh72-node1:~]# crm_resource --show-metadata stonith:fence_apc > fence_apc.xml
[root@rh72-node1:~]# jing /usr/share/cluster/relaxng/metadata.rng fence_apc.xml
/root/fence_apc.xml:194:40: error: attribute "timeout" not allowed here; expected attribute "automatic" or "on_target"
/root/fence_apc.xml:195:41: error: attribute "timeout" not allowed here; expected attribute "automatic" or "on_target"
[root@rh72-node1:~]# sed -n 190,200p fence_apc.xml
<action name="list"/>
<action name="list-status"/>
<action name="monitor"/>
<action name="metadata"/>
<action name="stop" timeout="20s"/>
<action name="start" timeout="20s"/>
</actions>
</resource-agent>
Unless there's a pressing reason to do this sooner, I'd rather defer this until the OCF standard has been updated upstream. We can bring up these differences as part of that discussion, and address whatever is left after that effort is done. As I see it this should be definitely tightly connected to the OCF standard update process which is currently ongoing upstream. There is no urgent need from pcs side. OCF update will not be in the 7.4 timeframe OCF standard update is not on the horizon at this moment, bumping again Sadly, the planned updates to the OCF standard have repeatedly been pushed back. An upstream bug has been filed for this issue, and this report will be closed. If the standard is updated, and developer time becomes available to work on the pacemaker side, we can reopen this. |
Description of problem: Some resource agents' metadata do not conform to the xml schema. It might seem the bug should have been filed againts resource agents. However, pacemaker provides metadata for lsb, systemd and service agents which do not conform either. Fence agents' metadata obtained from pacemaker do not conform either. Also various metadata defined and provided by pacemaker (stonithd, cib, cmrd, pengine) do not conform. Nagios agents metadata are the only one which are ok. When fixing the agents / pacemaker it is very likely to find out the metadata schema needs to be updated and extended to support new features already included in metadata. For now I filed the bug againts pacemaker. It can be cloned to other components if needed once investigated. Version-Release number of selected component (if applicable): pacemaker-1.1.15-10.el7 resource-agents-3.9.5-81.el7 fence-agents-all-4.0.11-27.el7 How reproducible: always, easily Steps to Reproduce: Output from the validator is really long, considering the number of agents, so I only put the commands here and ommit the actual output. For resource agents (I left out valid agents from the output): # cd /usr/share/resource-agents # for agent in `pcs resource list ocf: --nodesc`; do echo $agent; crm_resource --show-metadata $agent | xmllint --noout --valid --dtdvalid ra-api-1.dtd -; done Agents with not valid metadata are: ocf:heartbeat:portblock ocf:openstack:nova-compute-wait ocf:openstack:NovaEvacuate ocf:pacemaker:remote For lsb, service and systemd agents (most likely all agents in the given standard have the same errors): # cd /usr/share/resource-agents # for agent in `pcs resource list lsb: --nodesc`; do echo $agent; crm_resource --show-metadata $agent | xmllint --noout --valid --dtdvalid ra-api-1.dtd -; done # for agent in `pcs resource list systemd: --nodesc`; do echo $agent; crm_resource --show-metadata $agent | xmllint --noout --valid --dtdvalid ra-api-1.dtd -; done # for agent in `pcs resource list service: --nodesc`; do echo $agent; crm_resource --show-metadata $agent | xmllint --noout --valid --dtdvalid ra-api-1.dtd -; done This shows not a single agent's metadata is valid. For fence agents: # cd /usr/share/resource-agents # for agent in `pcs stonith list --nodesc`; do echo $agent; crm_resource --show-metadata stonith:$agent | xmllint --noout --valid --dtdvalid ra-api-1.dtd -; done This shows not a single agent's metadata is valid. Fence agents package however ships its own schema, let's try it: # for agent in `pcs stonith list --nodesc`; do echo $agent; crm_resource --show-metadata stonith:$agent | xmllint --noout --relaxng /usr/share/cluster/relaxng/metadata.rng -; done This again shows not a single agent's metadata is valid. fence_apc -:194: element action: Relax-NG validity error : Invalid attribute timeout for element action -:195: element action: Relax-NG validity error : Invalid attribute timeout for element action - fails to validate {snip} fence_xvm -:5: element parameters: Relax-NG validity error : Did not expect element parameters there - fails to validate Original metadata provided by agents mostly conform, though: # for agent in `pcs stonith list --nodesc`; do echo $agent; $agent -o metadata | xmllint --noout --relaxng /usr/share/cluster/relaxng/metadata.rng -; done fence_apc - validates {snip} fence_xvm -:4: element parameters: Relax-NG validity error : Did not expect element parameters there - fails to validate Metadata defined in pacemaker: # for agent in {cib,crmd,pengine,stonithd}; do echo $agent; /usr/libexec/pacemaker/$agent metadata | xmllint --noout --valid --dtdvalid ra-api-1.dtd -; done This shows not a single agent's metadata is valid. Actual results: Metadata do not conform to the schema. Expected results: Metadata do conform to the schema.