Cannot start cloudstack-agent if there is vm not managed by cloudstack on the host

[weizhouapache]

the agent.log shows

2023-10-03 12:18:23,692 ERROR [cloud.agent.AgentShell] (main:null) (logid:) Unable to start agent:
javax.naming.ConfigurationException: Error while creating Agent with class [com.cloud.hypervisor.kvm.resource.LibvirtComputingResource]. [Root exception is java.lang.NumberFormatException: null]
    at com.cloud.agent.AgentShell.launchAgentFromClassInfo(AgentShell.java:434)
    at com.cloud.agent.AgentShell.launchAgent(AgentShell.java:415)
    at com.cloud.agent.AgentShell.start(AgentShell.java:511)
    at com.cloud.agent.AgentShell.main(AgentShell.java:541)
Caused by: java.lang.NumberFormatException: null
    at java.base/java.lang.Integer.parseInt(Integer.java:614)
    at java.base/java.lang.Integer.parseInt(Integer.java:770)
    at com.cloud.hypervisor.kvm.resource.LibvirtDomainXMLParser.parseDomainXML(LibvirtDomainXMLParser.java:311)
    at com.cloud.hypervisor.kvm.resource.LibvirtComputingResource.setupMemoryBalloonStatsPeriod(LibvirtComputingResource.java:1301)
    at com.cloud.hypervisor.kvm.resource.LibvirtComputingResource.configure(LibvirtComputingResource.java:1244)
    at com.cloud.agent.Agent.<init>(Agent.java:190)
    at com.cloud.agent.AgentShell.launchNewAgent(AgentShell.java:452)
    at com.cloud.agent.AgentShell.launchAgentFromClassInfo(AgentShell.java:431)
    … 3 more

the issue is because there are 2 vms which are not created by cloudstack on the host. The xml of the vm does not have rate period and bytes in the rng secion. It should be

<rng model='virtio'>
<rate period='5000' bytes='4096' />
<backend model='random'>/dev/random</backend>
</rng>

Stopping the vms and edit the xml definition of vms should fix the issue.

ISSUE TYPE

• Bug Report

COMPONENT NAME

CLOUDSTACK VERSION

CONFIGURATION

OS / ENVIRONMENT

SUMMARY

STEPS TO REPRODUCE

EXPECTED RESULTS

ACTUAL RESULTS

[harikrishna-patnala]

@weizhouapache do we have to parse these VMs in first place ? or do we need to skip this section if does not exist for such VMs ?

[weizhouapache]

LibvirtComputingResource

I think option 2 is better.
we could skip the vms if the name does not start with i-/r-/s-/v- (not ideal but should work).

[DaanHoogland]

LibvirtComputingResource

I think option 2 is better. we could skip the vms if the name does not start with i-/r-/s-/v- (not ideal but should work).

couldn't the names start with those prefixes anyway (especially after (un)manage for kvm is merged)? Not sure how feasible it is to get information about that from the MS during this process. Maybe those are guaranteed to have the correct ,...

[weizhouapache]

LibvirtComputingResource

I think option 2 is better. we could skip the vms if the name does not start with i-/r-/s-/v- (not ideal but should work).

couldn't the names start with those prefixes anyway (especially after (un)manage for kvm is merged)? Not sure how feasible it is to get information about that from the MS during this process. Maybe those are guaranteed to have the correct ,...

@DaanHoogland
I think the challenge is, cloudstack agent cannot get the list of available vms from management server during start. there are few options
(1) determine if vm belongs to cloudstack by vm name (prefix), but it will not work if we import unmanaged vms in 4.19 as Daan said.
(2) determine if vm belongs to cloudstack by sysinfo, which were added in commit 3894d34

  <sysinfo type='smbios'>
    <system>
      <entry name='manufacturer'>Apache Software Foundation</entry>
      <entry name='product'>CloudStack KVM Hypervisor</entry>
      <entry name='uuid'>0820a0ca-f7e8-470a-a4f9-7a58fb75a6e4</entry>
    </system>
  </sysinfo>

(3) add a try-catch block and ignore the error so that cloudstack-agent can be started even if the vm xml cannot be parsed.

(4) move the setupMemoryBalloonStatsPeriod step out from start process (added in #6358).

option 3 looks like the most simple.

[harikrishna-patnala]

Yes agreed @weizhouapache I'm working on option 3

[DaanHoogland]

should be fixed by #8049