IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
co-authored-by: Pavel Czerný <pczerny@opennebula.systems>
co-authored-by: Frederick Borges <fborges@opennebula.io>
co-authored-by: Christian González <cgonzalez@opennebula.io>
* VNET updates trigger a driver action on running VMs with NICs in the
network.
* VNET includes a sets with VM status: updated, outdated, error and
updating. With VMs in each state.
* VNET flags error situations with a new state UPDATE_FAILURE.
* The same procedure is applied when an AR is updated (only VMs in that
AR are updated).
* A new options in the one.vn.recover API call enable to recover or
retry this VM update operations.
* The following attributes can be live-updated per VNET driver:
- PHYDEV (novlan, vlan, ovs driver)
- MTU (vlan, ovs driver)
- VLAN_ID (vlan, ovs driver)
- QINQ_TYPE (ovs driver)
- CVLANS (ovs driver)
- VLAN_TAGGED_ID (ovs driver)
- OUTER_VLAN_ID (ovs driver)
- INBOUND_AVG_BW (SG, ovs driver + KVM)
- INBOUND_PEAK_BW (SG, ovs driver + KVM)
- INBOUND_PEAK_KB (SG, ovs driver + KVM)
- OUTBOUND_AVG_BW (SG, ovs driver + KVM)
- OUTBOUND_PEAK_BW (SG, ovs driver + KVM)
- OUTBOUND_PEAK_KB (SG, ovs driver + KVM)
* New API call one.vm.updatenic, allows to update individual NICs
without the need of detach/attach (only QoS supported).
* Update operations for: 802.1Q, bridge, fw, ovswitch, ovswitch_vxlan
and vxlan network drivers.
* VNET attributes (old values) stored in VNET_UPDATE to allow
implementation of update operations. The attribute is removed after a
successful update.
* Updates to CLI onevnet (--retry option) / onevm (nicupdate command)
* XSD files updated to reflect the new data model
* Ruby and JAVA bindings updated: new VNET state and recover option, new
VM API call.
* Suntone and Fireedge implementation (lease status, recover option, new
states)
TODO: Virtual Functions does not support this functionality
iii
- Adds new configuration attribute MODE: FULL or INCREMENTAL for
BACKUP_CONFIG. FULL backups uses a differen backup image.
- INCREMENTAL backup information is together with the backup image.
Example:
<BACKUP_INCREMENTS>
<INCREMENT>
<DATE><![CDATA[1667770552]]></DATE>
<ID><![CDATA[0]]></ID>
<PARENT_ID><![CDATA[-1]]></PARENT_ID>
<SIZE><![CDATA[172]]></SIZE>
<SOURCE><![CDATA[bb828060]]></SOURCE>
<TYPE><![CDATA[FULL]]></TYPE>
</INCREMENT>
<INCREMENT>
<DATE><![CDATA[1667770604]]></DATE>
<ID><![CDATA[1]]></ID>
<PARENT_ID><![CDATA[0]]></PARENT_ID>
<SIZE><![CDATA[1]]></SIZE>
<SOURCE><![CDATA[ca0de5f6]]></SOURCE>
<TYPE><![CDATA[INCREMENT]]></TYPE>
</INCREMENT>
<INCREMENT>
<DATE><![CDATA[1667770700]]></DATE>
<ID><![CDATA[2]]></ID>
<PARENT_ID><![CDATA[1]]></PARENT_ID>
<SIZE><![CDATA[39]]></SIZE>
<SOURCE><![CDATA[e9897d6a]]></SOURCE>
<TYPE><![CDATA[INCREMENT]]></TYPE>
</INCREMENT>
</BACKUP_INCREMENTS>
This information only appears on incremental backups
- Internal BACKUP_CONFIG data includes information about the current
active backup and the last increment id.
- Backup operation includes a new parameter: reset. This "closes" the
current active incremental chain and creates a new FULL backup.
- Backup drivers store backups with increment index (0 = FULL) e.g.
disk.0.0.
- Incremental backups are only allowed for VMs using all disks in qcow2
format.
- Backup configuration cannot be changed while doing a VM backup.
- Downloader strings includes backup chains <inc_id>:<backup_ref>,...
- Restic downloader has been updated to support backup chains. Disk
images are rebased across increments.
co-authored-by: Frederick Borges <fborges@opennebula.io>
co-authored-by: Neal Hansen <nhansen@opennebula.io>
co-authored-by: Daniel Clavijo Coca <dclavijo@opennebula.io>
co-authored-by: Pavel Czerný <pczerny@opennebula.systems>
BACKUP INTERFACE
=================
* Backups are exposed through a a special Datastore (BACKUP_DS) and
Image (BACKUP) types. These new types can only be used for backup'ing
up VMs. This approach allows to:
- Implement tier based backup policies (backups made on different
locations).
- Leverage access control and quota systems
- Support differnt storage and backup technologies
* Backup interface for the VMs:
- VM configures backups with BACKUP_CONFIG. This attribute can be set
in the VM template or updated with updateconf API call. It can include:
+ BACKUP_VOLATILE: To backup or not volatile disks
+ FS_FREEZE: How the FS is freeze for running VMs (qemu-agent,
suspend or none). When possible backups are crash consistent.
+ KEEP_LAST: keep only a given number of backups.
- Backups are initiated by the one.vm.backup API call that requires
the target Datastore to perform the backup (one-shot). This is
exposed by the onevm backup command.
- Backups can be periodic through scheduled actions.
- Backup configuration is updated with one.vm.updateconf API call.
* Restore interface:
- Restores are initiated by the one.image.restore API call. This is
exposed by oneimage restore command.
- Restore include configurable options for the VM template
+ NO_IP: to not preserve IP addresses (but keep the NICs and network
mapping)
+ NO_NIC: to not preserve network mappings
- Other template attributes:
+ Clean PCI devices, including network configuration in case of TYPE=NIC
attributes. By default it removes SHORT_ADDRESS and leave the "auto"
selection attributes.
+ Clean NUMA_NODE, removes node id and cpu sets. It keeps the NUMA node
- It is possible to restore single files stored in the repository by
using the backup specific URL.
* Sunstone (Ruby version) has been updated to expose this feautres.
BACKUP DRIVERS & IMPLEMENTATION
===============================
* Backup operation is implemented by a combination of 3 driver operations:
- VMM. New (internal oned <-> one_vmm_exec.rb) to orchestrate
backups for RUNNING VMs.
- TM. This commit introduces 2 new operations (and their
corresponding _live variants):
+ pre_backup(_live): Prepares the disks to be back'ed up in the
repository. It is specific to the driver: (i) ceph uses the export
operation; (ii) qcow2/raw uses snapshot-create-as and fs_freeze as
needed.
+ post_backup(_live): Performs cleanning operations, i.e. KVM
snapshots or tmp dirs.
- DATASTORE. Each backup technology is represented by its
corresponfing driver, that needs to implement:
+ backup: it takes the VM disks in file (qcow2) format and stores it
the backup repository.
+ restore: it takes a backup image and restores the associated disks
and VM template.
+ monitor: to gather available space in the repository
+ rm: to remove existing backups
+ stat: to return the "restored" size of a disk stored in a backup
+ downloader pseudo-URL handler: in the form
<backup_proto>://<driver_snapshot_id>/<disk filename>
BACKUP MANAGEMENT
=================
Backup actions may potentially take some time, leaving some vmm_exec threads in
use for a long time, stucking other vmm operations. Backups are planned
by the scheduler through the sched action interface.
Two attributes has been added to sched.conf:
* MAX_BACKUPS max active backup operations in the cloud. No more
backups will be started beyond this limit.
* MAX_BACKUPS_HOST max number of backups per host
* Fix onevm CLI to properly show and manage schedule actions. --schedule
supports now, as well as relative times +<seconds_from_stime>
onvm backup --schedule now -d 100 63
* Backup is added as VM_ADMIN_ACTIONS in oned.conf. Regular users needs
to use the batch interface or request specific permissions
Internal restructure of Scheduler:
- All sched_actions interface is now in SchedActionsXML class and files.
This class uses references to VM XML, and MUST be used in the same
lifetime scope.
- XMLRPC API calls for sched actions has been moved to ScheduledActionXML.cc as
static functions.
- VirtualMachineActionPool includes counters for active backups (total
and per host).
SUPPORTED PLATFORMS
====================
* hypervisor: KVM
* TM: qcow2/shared/ssh, ceph
* backup: restic, rsync
Notes on Ceph
* Ceph backups are performed in the following steps:
1. A snapshot of each disk is taken (group snapshots cannot be used as
it seems we cannot export the disks afterwards)
2. Disks are export to a file
3. File is converted to qcow2 format
4. Disk files are upload to the backup repo
TODO:
* Confirm crash consistent snapshots cannot be used in Ceph
TODO:
* Check if using VM dir instead of full path is better to accomodate
DS migrations i.e.:
- Current path: /var/lib/one/datastores/100/53/backup/disk.0
- Proposal: 53/backup/disk.0
RESTIC DRIVER
=============
Developed together with this feature is part of the EE edtion.
* It supports the SFTP protocol, the following attributes are
supported:
- RESTIC_SFTP_SERVER
- RESTIC_SFTP_USER: only if different from oneadmin
- RESTIC_PASSWORD
- RESTIC_IONICE: Run restic under a given ionice priority (class 2)
- RESTIC_NICE: Run restic under a given nice
- RESTIC_BWLIMIT: Limit restic upload/download BW
- RESTIC_COMPRESSION: Restic 0.14 implements compression (three modes:
off, auto, max). This requires repositories version 2. By default,
auto is used (average compression without to much CPU usage)
- RESTIC_CONNECTIONS: Sets the number of concurrent connections to a
backend (5 by default). For high-latency backends this number can be
increased.
* downloader URL: restic://<datastore_id>/<snapshot_id>/<file_name>
snapshot_id is the restic snapshot hash. To recover single disk images
from a backup. This URLs support:
- RESTIC_CONNECTIONS
- RESTIC_BWLIMIT
- RESTIC_IONICE
- RESTIC_NICE
These options needs to be defined in the associated datastore.
RSYNC DRIVER
=============
A rsync driver is included as part of the CE distribution. It uses the
rsync tool to store backups in a remote server through SSH:
* The following attributes are supported to configure the backup
datastore:
- RSYNC_HOST
- RSYNC_USER
- RSYNC_ARGS: Arguments to perform the rsync operatin (-aS by default)
* downloader URL: rsync://<ds_id>/<vmid>/<hash>/<file> can be used to recover
single files from an existing backup. (RSYNC_HOST and RSYN_USER needs
to be set in ds_id
EMULATOR_CPUS
=============
This commit includes a non related backup feature:
* Add EMULATOR_CPUS (KVM). This host (or cluster attribute) defines the
CPU IDs where the emulator threads will be pinned. If this value is
not defined the allocated CPU wll be used when using a PIN policy.
(cherry picked from commit a9e6a8e000e9a5a2f56f80ce622ad9ffc9fa032b)
F OpenNebula/one#5516: adding rsync backup driver
(cherry picked from commit fb52edf5d009dc02b071063afb97c6519b9e8305)
F OpenNebula/one#5516: update install.sh, add vmid to source, some polish
Signed-off-by: Neal Hansen <nhansen@opennebula.io>
(cherry picked from commit 6fc6f8a67e435f7f92d5c40fdc3d1c825ab5581d)
F OpenNebula/one#5516: cleanup
Signed-off-by: Neal Hansen <nhansen@opennebula.io>
(cherry picked from commit 12f4333b833f23098142cd4762eb9e6c505e1340)
F OpenNebula/one#5516: update downloader, default args, size check
Signed-off-by: Neal Hansen <nhansen@opennebula.io>
(cherry picked from commit 510124ef2780a4e2e8c3d128c9a42945be38a305)
LL
(cherry picked from commit d4fcd134dc293f2b862086936db4d552792539fa)
- SHORT_ADDRESS can be used to select specific devices (useful on
homogenous clusters or nic attach operations). This name has been
selected because of:
1. It is the attrbute shown host info
2. It does not collide with the reserved ADDRESS attribute
- New test and add functions that considers both allocation methods: by
name (VENDOR/CLASS/DEVICE) or address (SHORT_ADDRESS)
- Parameter check on VM creation
- revert and add method use the same pci_attribute function to add info
to the VM PCI attribute
- Remove well-known attributes when parsing PCI devices (ADDRESS,
PREV_ADDRES, BUS, FUNCTION, SLOT, NUMA_NODE, UUID)
- Support for attach and detach NIC with PCI attributes
* onevm_exec.rb looks for PCI devices for ATTACH=YES when attaching/detaching an interface
* script action are now written in Ruby
* KVM module with common actions (hostdev/interface device str)
* Minor changes in xmlparser and OpenNebulaVM classes
- PCI selection options to onevm nic-attach:
* pci short_address
* pci_device device ID
* pci_vendor vendor ID
* pci_class class ID
- VF can be configured by setting some parameters through IP link (e.g.
MAC or VLAN_ID). This commit includes a mixin to activate_vf
* one_vmm_exec.rb looks for PCI VN_MAD drivers
* VM class (VNM) adds a @pcis array
* activate_vf should be called in the pre stage. The following drivers
acticate VFs (VLAN_ID is implemented as 802.1Q tag)
- 802.1Q
- bridge
- fw
- ovswitch
* Supported attributes for VF:
- MAC
- VLAN_ID
- spoof checking (SPOOFCHK)
- trust (TRUST)
- Predictable PCI addresses for guests is only enabled if PCI bus 1
is present (PCI bridge bus cannot be hotplugged)
- Improve integration with Libvirt/QEMU:
* When attach, only activate the VF being attached
* Attach: Use <interface> and not <hostdev> for VF. There seems to be a race
condition between accessing the vfio device and permission setup.
* Attach: Remove address on attach as it may fail beacuse PCI controller
is not present, e.g.:
ATTACHNIC: Could not attach NIC to 28534240: error: Failed to attach device
from /dev/fd/63 error: XML error: Invalid PCI address 0000:01:01.0.
Only PCI buses up to 0 are available ExitCode: 1
* Detach: Detach always use <hostdev> as libvirt fails to identified the
device just by address when using <interface>
- Suntone Support: The "Hardware" profile of a network interface (NIC) of a VM can be of three types, now exposed in Sunstone:
* "Emulated" it includes the hardware model emulated by Qemu
* "PCI - Automatic" oned hw scheduler will pick the best PCI device for the NIC
* "PCI - Manual" user can specify the PCI device by its short-address as shown in host information
This commits also enables the attach/detach operations on PCI based NICs in Suntone. *Note*: only for KVM VMs
(reverts commit aafd5f3014a29266141cef9dd89d79a6323eda44)
(reverts commit 671ef0b2d91aba00b1e6c63e19c4a50c3af20f6f)
- SHORT_ADDRESS can be used to select specific devices (useful on
homogenous clusters or nic attach operations). This name has been
selected because of:
1. It is the attrbute shown host info
2. It does not collide with the reserved ADDRESS attribute
- New test and add functions that considers both allocation methods: by
name (VENDOR/CLASS/DEVICE) or address (SHORT_ADDRESS)
- Parameter check on VM creation
- revert and add method use the same pci_attribute function to add info
to the VM PCI attribute
- Remove well-known attributes when parsing PCI devices (ADDRESS,
PREV_ADDRES, BUS, FUNCTION, SLOT, NUMA_NODE, UUID)
- Support for attach and detach NIC with PCI attributes
* onevm_exec.rb looks for PCI devices for ATTACH=YES when attaching/detaching an interface
* script action are now written in Ruby
* KVM module with common actions (hostdev/interface device str)
* Minor changes in xmlparser and OpenNebulaVM classes
- PCI selection options to onevm nic-attach:
* pci short_address
* pci_device device ID
* pci_vendor vendor ID
* pci_class class ID
- VF can be configured by setting some parameters through IP link (e.g.
MAC or VLAN_ID). This commit includes a mixin to activate_vf.
* one_vmm_exec.rb looks for PCI VN_MAD drivers
* VM class (VNM) adds a @pcis array
* activate_vf should be called in the pre stage. The following drivers
acticate VFs (VLAN_ID is implemented as 802.1Q tag)
- 802.1Q
- bridge
- fw
- ovswitch
- Improve integration with Libvirt/QEMU:
* When attach, only activate the VF being attached
* Attach: Use <interface> and not <hostdev> for VF. There seems to be a race
condition between accessing the vfio device and permission setup.
* Attach: Remove address on attach as it may fail beacuse PCI controller
is not present, e.g.:
ATTACHNIC: Could not attach NIC to 28534240: error: Failed to attach device
from /dev/fd/63 error: XML error: Invalid PCI address 0000:01:01.0.
Only PCI buses up to 0 are available ExitCode: 1
* Detach: Detach always use <hostdev> as libvirt fails to identified the
device just by address when using <interface>
* VNETs include state to flag their configuration status
(READY/ERROR/LOCK)
* Hooks can be program based on VNET state transitions
* Include Java/Go/Ruby bindings
* New vCenter network drivers
* Place-holder to customize actions (create/delete) for other netowrk drivers
* Sunstone integration
* OneFlow syncs flow creation on VNET status
* Remove deprecated vcenter hooks and autoregistration
co-authored-by: Frederick Ernesto Borges Noronha <fborges@opennebula.io>
co-authored-by: Alejandro Huertas <ahuertas@opennebula.io>
co-authored-by: Pavel Czerny <pczerny@opennebula.systems>
* New probe to get CGROUP_VERSION to the host
* LibvirtKVM driver set shares based on version (1024/100)
* Limits for min and max are check and enforced
SchedActions are now a complete managed type. It is in in /TEMPLATE and
not /USER_TEMPLATE and includes dedicated API Calls to manage the
entries:
- one.vm.schedadd (add new sched_action)
- one.vm.schedupdate (update an exiting sched_action)
- one.vm.scheddelete (delete an action)
The scheduler now use this API. This will prevent race conditions
between update operations from different processes.
co-authored-by: Pavel Czerný <pczerny@opennebula.systems>
Includes the following commits from f-5353:
- 62d3d38d7fb135b82874a5076a7f06c23675e0b7
- 52f3f066603c478061891175c008a58cfed3dc43
- 0e82f9e5471c3cf0e21969393299279c6e739138
- 8c32050940f528ea876e8293e6fbbf56a373bd54
