Cloudstack assigns Volume to wrong host

[HeinzM]

Problem:

1. Deploy Instances with terraform
2. Cloudstack creates Instances and Volumes
3. Volumes were mapped to wrong host, even to hosts of different cluster ( Primary Storage is set to cluster scope)

Setup:
CloudStack 4.21.0
OS: Rocky Linux 9.6
Hypervisor: KVM/QEMU
Primary Storage: Powerflex 4.6.1.0 (715)
Secondary Storage: NFS

Infrastruce-Setup:
1 Zone
2 Pods with 1 Cluster each
4 Hosts in each Cluster

example:

Currently, a working state can occasionally be achieved with very frequent deployment runs.
Most of the time, the mapping is not executed correctly.

[boring-cyborg[bot]]

Thanks for opening your first issue here! Be sure to follow the issue template!

[HeinzM]

We created a new Deployment.
VR with ID 330
Instance 1 with ID 329
and Instance 2 with ID 331.
The Volumes are named correctly in CloudStack.

but in Powerflex the following volumes appear:

[HeinzM]

Is there any connection with the missing host_ip entries in the database?

[sureshanaparti]

Hi @HeinzM There is no connection with the host_ip entries for the PowerFlex volumes. Are the PowerFlex storage pools (i.e. primary storages with cluster scope) added in each CloudStack cluster belongs to same or different PowerFlex storage cluster?

pool_id column in cloud.volumes table maps to the current storage pool of the volume, and storage host details are in cloud.storage_pool table. Can you query with the following sql:

SELECT vol.name, vol.pool_id, vol.last_pool_id, vol.instance_id, vol.device_id, vol.size, vol.path, pool.name AS pool_name, pool.cluster_id AS pool_cluster_id, pool.host_address AS pool_host_address, pool.path AS pool_path FROM cloud.volumes vol JOIN cloud.storage_pool pool ON vol.pool_id = pool.id WHERE vol.instance_id IN (329, 330, 331);

[HeinzM]

Hello @sureshanaparti,
there is one specific storage system per cluster. cluster 1 gets powerflex system 1 and cluster 2 gets powerflex system 2.
The pfx systems are individual clusters
The query gets the following result:

[sureshanaparti]

Hello @sureshanaparti, there is one specific storage system per cluster. cluster 1 gets powerflex system 1 and cluster 2 gets powerflex system 2. The pfx systems are individual clusters The query gets the following result:

@HeinzM I don't issue any issue with storage pool host assignment for all these volumes. If your concern is regarding the host_ip column in cloud.volumes table, it is not applicable for PowerFlex volumes. The PowerFlex volume mapping is based on the host_id column in the vm_instance table (for the volumes attached to the instance, check instance_id column in the cloud.volumes table).

[HeinzM]

I'm confused right now. I expected that the volume name in cloudstack matches the volume descriptor in powerflex.

[sureshanaparti]

@HeinzM You can check the volume naming conventions here:
https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=158873968#PowerFlex/ScaleIOstorageplugin-NamingconventionsforScaleIOvolumes

[sureshanaparti]

I'm confused right now. I expected that the volume name in cloudstack matches the volume descriptor in powerflex.

@HeinzM Is it clear now? Are you looking for any other information?

[HeinzM]

It's clear now.
I'm looking for information regarding another topic. Is it possible to encrypt network traffic in guest networks between instances if the running software itself cannot do so? Can you answer that here or shall I create a new discussion?