* kvm: Add support for cgroupv2 (#8252)
1. Problem description
In Apache CloudStack (ACS), when a VM is deployed in a host with the KVM hypervisor, an XML file is created in the assigned host, which has a property shares that defines the weight of the VM to access the host CPU. The value of this property has no unit, and it is a relative measure to calculate how much CPU a given VM will have in the host. However, this value has a limit, which depends on the version of cgroup utilized by the host's kernel. The problem lies at the range value of shares that varies between both versions: [2, 264144] for cgroups version 1; and [1, 10000] for cgroups version 2. Currently, ACS calculates the value of shares using Equation 1, presented below, where CPU is the number of cores and speed is the CPU frequency; both specified in the VM's compute offering. Therefore, if a compute offering has, for example, 6 cores at 2 GHz, the shares value will be 12000 and an exception will be thrown by libvirt if the host utilizes cgroup v2. The second version is becoming the default one in current Linux distributions; thus, it is necessary to address this limitation.
Equation 1
shares = CPU * speed
Fixes: #6744
2. Proposed changes
To address the problem described, we propose to apply a scale conversion considering the max shares of the host. Using the same formula currently utilized by ACS, it is possible to calculate the maximum shares of a VM for a given host. In other words, using the number of cores and the nominal speed of the host's CPU as the upper limit of shares allowed to a VM. Then, this value will be scaled to the allowed interval of [1, 10000] of cgroup v2 by using a linear scale conversion.
The VM shares would be calculated as Equation 2, presented below, where VM requested shares is the requested shares value calculated using Equation 1, cgroup upper limit is fixed with a value of 10000 (cgroups v2 upper limit), and host max shares is the maximum shares value of the host, calculated using Equation 1. Using Equation 2, the only case where a VM passes the cgroup v2 limit is when the user requests more resources than the host has, which is not possible with the current implementation of ACS.
Equation 2
shares = (VM requested shares * cgroup upper limit)/host max shares
To implement the proposal, the following APIs will be updated: deployVirtualMachine, migrateVirtualMachine and scaleVirtualMachine. When a VM is being deployed, a new verification will be added to find a suitable host. The max shares of each host will be calculated, and the VM calculated shares will be verified if it does not surpass the host's value. Likewise, the migration of VMs will have a similar new verification. Lastly, the scale of VMs will also have the same verification for the VM's host.
To determine the max shares of a given host, we will use the same equation currently used in ACS for calculating the shares of VMs, presented in Section 1. When Equation 1 is used to determine the maximum shares of a host, CPU is the number of cores of the host, and speed is the nominal CPU speed, i.e., considering the CPU's base frequency.
It is important to note that these changes are only for hosts with the KVM hypervisor using cgroup v2 for now.
* Update overcommit ratio during live VM migration
* minor refactoring
---------
Co-authored-by: Bryan Lima <42067040+BryanMLima@users.noreply.github.com>
* Introduced a new API "checkVolumeAndRepair" that allows users or admins to check and repair if any leaks observed.
Currently this is supported only for KVM
* some fixes
* Added unit tests
* addressed review comments
* add repair volume while granting access
* Changed repair parameter to accept both leaks/all values
* Introduced new global setting volume.check.and.repair.before.use to do volume check and repair before VM start or volume attach operations
* Added volume check and repair changes only during VM start and volume attach operations
* Refactored the names to look similar across the code
* Some code fixes
* remove unused code
* Renamed repair values
* Addressed review comments
* code refactored
* used volume name in logs
* Changed the API to Async and the setting scope to storage pool
* Fixed exit value handling with check volume command
* Fixed storage scope to the setting
* Fixed volume format issues
* Refactored the log messages
* Fix formatting
* Improve logs
* Remove unnecessary comments
* Use diamond inference
* Fix some logs
* Remove unnecessary unboxing
* Create method to handle job result
* Remove unused vars and fix some logics
* Extract code to method and few adjusts
* Use CollectionUtils
* Extract pending work job validation to method
* Create new constructors
* Extract work job and info creation to a method
* Extract submit async job to a method
* Extract find vm by id to a method
* Change log level from trace to debug
* Remove unnused methods and add logs
* Undo code remotion
* Remove asserts and fix conditionals
* Address @GabrielBrascher reviews
* Remove double quotes from keys in manual json
* Undo code remotion
* Add object to log
* Remove statement from try/catch
* Implement toString with ReflectionToStringBuilderUtils
* Fix errors related to merge main
Co-authored-by: Daniel Augusto Veronezi Salvador <daniel@scclouds.com.br>
Added support for PowerFlex/ScaleIO (v3.5 onwards) storage pool as a primary storage in CloudStack (for KVM hypervisor) and enabled VM/Volume operations on that pool (using pool tag).
Please find more details in the FS here:
https://cwiki.apache.org/confluence/x/cDl4CQ
Documentation PR: apache/cloudstack-documentation#169
This enables support for PowerFlex/ScaleIO (v3.5 onwards) storage pool as a primary storage in CloudStack
Other improvements addressed in addition to PowerFlex/ScaleIO support:
- Added support for config drives in host cache for KVM
=> Changed configuration "vm.configdrive.primarypool.enabled" scope from Global to Zone level
=> Introduced new zone level configuration "vm.configdrive.force.host.cache.use" (default: false) to force host cache for config drives
=> Introduced new zone level configuration "vm.configdrive.use.host.cache.on.unsupported.pool" (default: true) to use host cache for config drives when storage pool doesn't support config drive
=> Added new parameter "host.cache.location" (default: /var/cache/cloud) in KVM agent.properties for specifying the host cache path and create config drives on the "/config" directory on the host cache path
=> Maintain the config drive location and use it when required on any config drive operation (migrate, delete)
- Detect virtual size from the template URL while registering direct download qcow2 (of KVM hypervisor) templates
- Updated full deployment destination for preparing the network(s) on VM start
- Propagate the direct download certificates uploaded to the newly added KVM hosts
- Discover the template size for direct download templates using any available host from the zones specified on template registration
=> When zones are not specified while registering template, template size discovery is performed using any available host, which is picked up randomly from one of the available zones
- Release the VM resources when VM is sync-ed to Stopped state on PowerReportMissing (after graceful period)
- Retry VM deployment/start when the host cannot grant access to volume/template
- Mark never-used or downloaded templates as Destroyed on deletion, without sending any DeleteCommand
=> Do not trigger any DeleteCommand for never-used or downloaded templates as these doesn't exist and cannot be deleted from the datastore
- Check the router filesystem is writable or not, before performing health checks
=> Introduce a new test "filesystem.writable.test" to check the filesystem is writable or not
=> The router health checks keeps the config info at "/var/cache/cloud" and updates the monitor results at "/root" for health checks, both are different partitions. So, test at both the locations.
=> Added new script: "filesystem_writable_check.py" at /opt/cloud/bin/ to check the filesystem is writable or not
- Fixed NPE issue, template is null for DATA disks. Copy template to target storage for ROOT disk (with template id), skip DATA disk(s)
* Addressed some issues for few operations on PowerFlex storage pool.
- Updated migration volume operation to sync the status and wait for migration to complete.
- Updated VM Snapshot naming, for uniqueness in ScaleIO volume name when more than one volume exists in the VM.
- Added sync lock while spooling managed storage template before volume creation from the template (non-direct download).
- Updated resize volume error message string.
- Blocked the below operations on PowerFlex storage pool:
-> Extract Volume
-> Create Snapshot for VMSnapshot
* Added the PowerFlex/ScaleIO client connection pool to manage the ScaleIO gateway clients, which uses a single gateway client per Powerflex/ScaleIO storage pool and renews it when the session token expires.
- The token is valid for 8 hours from the time it was created, unless there has been no activity for 10 minutes.
Reference: https://cpsdocs.dellemc.com/bundle/PF_REST_API_RG/page/GUID-92430F19-9F44-42B6-B898-87D5307AE59B.html
Other fixes included:
- Fail the VM deployment when the host specified in the deployVirtualMachine cmd is not in the right state (i.e. either Resource State is not Enabled or Status is not Up)
- Use the physical file size of the template to check the free space availability on the host, while downloading the direct download templates.
- Perform basic tests (for connectivity and file system) on router before updating the health check config data
=> Validate the basic tests (connectivity and file system check) on router
=> Cleanup the health check results when router is destroyed
* Updated PowerFlex/ScaleIO storage plugin version to 4.16.0.0
* UI Changes to support storage plugin for PowerFlex/ScaleIO storage pool.
- PowerFlex pool URL generated from the UI inputs(Gateway, Username, Password, Storage Pool) when adding "PowerFlex" Primary Storage
- Updated protocol to "custom" for PowerFlex provider
- Allow VM Snapshot for stopped VM on KVM hypervisor and PowerFlex/ScaleIO storage pool
and Minor improvements in PowerFlex/ScaleIO storage plugin code
* Added support for PowerFlex/ScaleIO volume migration across different PowerFlex storage instances.
- findStoragePoolsForMigration API returns PowerFlex pool(s) of different instance as suitable pool(s), for volume(s) on PowerFlex storage pool.
- Volume(s) with snapshots are not allowed to migrate to different PowerFlex instance.
- Volume(s) of running VM are not allowed to migrate to other PowerFlex storage pools.
- Volume migration from PowerFlex pool to Non-PowerFlex pool, and vice versa are not supported.
* Fixed change service offering smoke tests in test_service_offerings.py, test_vm_snapshots.py
* Added the PowerFlex/ScaleIO volume/snapshot name to the paths of respective CloudStack resources (Templates, Volumes, Snapshots and VM Snapshots)
* Added new response parameter “supportsStorageSnapshot” (true/false) to volume response, and Updated UI to hide the async backup option while taking snapshot for volume(s) with storage snapshot support.
* Fix to remove the duplicate zone wide pools listed while finding storage pools for migration
* Updated PowerFlex/ScaleIO volume migration checks and rollback migration on failure
* Fixed the PowerFlex/ScaleIO volume name inconsistency issue in the volume path after migration, due to rename failure
* [CLOUDSTACK-10323] Allow changing disk offering during volume migration
This is a continuation of work developed on PR #2425 (CLOUDSTACK-10240), which provided root admins an override mechanism to move volumes between storage systems types (local/shared) even when the disk offering would not allow such operation. To complete the work, we will now provide a way for administrators to enter a new disk offering that can reflect the new placement of the volume. We will add an extra parameter to allow the root admin inform a new disk offering for the volume. Therefore, when the volume is being migrated, it will be possible to replace the disk offering to reflect the new placement of the volume.
The API method will have the following parameters:
* storageid (required)
* volumeid (required)
* livemigrate(optional)
* newdiskofferingid (optional) – this is the new parameter
The expected behavior is the following:
* If “newdiskofferingid” is not provided the current behavior is maintained. Override mechanism will also keep working as we have seen so far.
* If the “newdiskofferingid” is provided by the admin, we will execute the following checks
** new disk offering mode (local/shared) must match the target storage mode. If it does not match, an exception will be thrown and the operator will receive a message indicating the problem.
** we will check if the new disk offering tags match the target storage tags. If it does not match, an exception will be thrown and the operator will receive a message indicating the problem.
** check if the target storage has the capacity for the new volume. If it does not have enough space, then an exception is thrown and the operator will receive a message indicating the problem.
** check if the size of the volume is the same as the size of the new disk offering. If it is not the same, we will ALLOW the change of the service offering, and a warning message will be logged.
We execute the change of the Disk offering as soon as the migration of the volume finishes. Therefore, if an error happens during the migration and the volume remains in the original storage system, the disk offering will keep reflecting this situation.
* Code formatting
* Adding a test to cover migration with new disk offering (#4)
* Adding a test to cover migration with new disk offering
* Update test_volumes.py
* Update test_volumes.py
* fix test_11_migrate_volume_and_change_offering
* Fix typo in Java doc
Remove maven standard module (which only a few were using) and get ride of maven customization for the projects structure.
- moved all directories to src/main/java, src/main/resources, src/main/scripts, src/test/java, src/test/resources
- grep scan to search for src/com and src/org left over
- grep for <project>/scripts to fix pom.xml configuration
- remove custom <build> configuration in pom.xml
Signed-off-by: Marc-Aurèle Brothier <m@brothier.org>
Vishesh
Harikrishna
Suresh Kumar Anaparti
João Jandre
Daniel Augusto Veronezi Salvador
sureshanaparti
Rohit Yadav
Rafael Weingärtner
Marc-Aurèle Brothier