CloudStack 4.22: `prepareHostForMaintenance` throws NPE when stale destroyed volume references removed storage pool

从第 7558 行的 UserVmManagerImpl.isAnyVmVolumeUsingLocalStorage 开始，跟踪其在 isVMUsingLocalStorage 和 ResourceManagerImpl.doMaintain 中的调用方。重现已销毁的陈旧卷和已移除的存储池状态，然后验证维护不再引发未处理的 NullPointerException，并确认针对陈旧卷所选择的行为已由相关测试覆盖。

problem

Description

While attempting to place a KVM host into Maintenance mode in CloudStack 4.22, the maintenance operation failed with a NullPointerException.

The issue appears to occur when a VM has a stale/destroyed volume entry in the volumes table that still references a removed storage pool.

Instead of gracefully ignoring the stale volume metadata or returning a user-facing validation error, CloudStack crashes during maintenance preparation.

Environment

• CloudStack version: 4.22
• Hypervisor: KVM
• Primary storage: NFS
• Database: MySQL/MariaDB

Steps to reproduce

1. Have a VM with:

   • one valid active ROOT volume
   • one stale/destroyed ROOT volume entry still present in the volumes table

2. The stale volume references a storage pool which has already been removed.

3. Attempt to place the host running the VM into Maintenance mode.

Example problematic DB state:

VM

SELECT id, uuid, name, instance_id, pool_id, state, removed
FROM volumes
WHERE instance_id = 446
ORDER BY id;

Result:

+------+--------------------------------------+----------+-------------+---------+---------+---------+
| id   | uuid                                 | name     | instance_id | pool_id | state   | removed |
+------+--------------------------------------+----------+-------------+---------+---------+---------+
|  928 | 6dea3d6f-bd6d-4e8b-9524-6e99c029694c | ROOT-446 |         446 |       4 | Destroy | NULL    |
| 1554 | 80acf9ae-b047-41b8-bded-cdceb6de7051 | ROOT-446 |         446 |       2 | Ready   | NULL    |
+------+--------------------------------------+----------+-------------+---------+---------+---------+

Storage pool state

The stale volume references storage pool ID 4, which is already removed:

storage_pool_name: Export-Domain
storage_pool_removed: 2026-04-23 13:44:08

Actual result

Host maintenance fails with:

java.lang.NullPointerException: Cannot invoke
"org.apache.cloudstack.storage.datastore.db.StoragePoolVO.isLocal()"
because "storagePool" is null

Relevant stack trace:

at com.cloud.vm.UserVmManagerImpl.isAnyVmVolumeUsingLocalStorage(UserVmManagerImpl.java:7558)
at com.cloud.vm.UserVmManagerImpl.isVMUsingLocalStorage(UserVmManagerImpl.java:7121)
at com.cloud.resource.ResourceManagerImpl.doMaintain(ResourceManagerImpl.java:1553)
at com.cloud.resource.ResourceManagerImpl.maintain(ResourceManagerImpl.java:1653)
at org.apache.cloudstack.api.command.admin.host.PrepareForHostMaintenanceCmd.execute(PrepareForHostMaintenanceCmd.java:99)

Expected result

CloudStack should not throw an unhandled NullPointerException.

Possible expected behavior:

• ignore destroyed/removed stale volumes during maintenance evaluation
• skip volumes attached to removed pools
• or return a proper validation error identifying the problematic VM/volume

Workaround

Marking the stale destroyed volume row as removed allowed maintenance to proceed:

UPDATE volumes
SET removed = NOW()
WHERE id = 928
  AND instance_id = 446
  AND state = 'Destroy'
  AND removed IS NULL
  AND pool_id = 4;

Additional notes

The issue appears to be triggered specifically by:

• stale destroyed volume rows
• still linked to an active/running VM
• referencing removed storage pools
• while evaluating VM local-storage usage during host maintenance

versions

The versions ACS 4.22, KVM (should not be relevant)

The steps to reproduce the bug

...

What to do about it?

No response