* Introducing Storage Access Groups to define the host and storage pool connections
In CloudStack, when a primary storage is added at the Zone or Cluster scope, it is by default connected to all hosts within that scope. This default behavior can be refined using storage access groups, which allow operators to control and limit which hosts can access specific storage pools.
Storage access groups can be assigned to hosts, clusters, pods, zones, and primary storage pools. When a storage access group is set on a cluster/pod/zone, all hosts within that scope inherit the group. Connectivity between a host and a storage pool is then governed by whether they share the same storage access group.
A storage pool with a storage access group will connect only to hosts that have the same storage access group. A storage pool without a storage access group will connect to all hosts, including those with or without a storage access group.
Sometimes hypervisor hosts (direct agents) stuck with Disconnect state during agent rebalancing activity across multiple management server nodes. This issue was noticed during frequent restart of the management server nodes in the cluster.
When there are multiple management server nodes in a cluster, if one or more nodes are shutdown/start/restart, CloudStack will rebalance the hosts among the remaining nodes or move the nodes to the newly joined management server nodes. During the rebalancing period multiple operations could happen including:
- DirectAgentScan at interval of configured direct.agent.scan.interval
- AgentRebalanceScan to identify and schedule rebalance agents
- TransferAgentScan to transfer the host from original owner to future owner
**Current Rebalance behavior**
1. For hosts that have AgentAttache && not forForward but in Disconnect state, CloudStack simply ignore these hosts without trying to ping again or update the status of the host.
2. For hosts that have AgentAttache && forForward, CloudStack removes the agent but still try to loadDirectlyConnectedHost.
**Improved Rebalance behavior**
During DirectAgentScan: scanDirectAgentToLoad(), identify hosts that for self-managed hosts that are in Disconnect state (disconnected after pingtimeout).
1. For hosts that have AgentAttache and is forForward, CloudStack should remove the agent
2. For hosts that have AgentAttache and is not forForward but in Disconnect state, CloudStack should try to investigate and update the status to Up if host is pingable.
3. For hosts that don't have AgentAttache, CloudStack should try to loadDirectlyConnectedHost.
* KVM incremental snapshot feature
* fix log
* fix merge issues
* fix creation of folder
* fix snapshot update
* Check for hypervisor type during parent search
* fix some small bugs
* fix tests
* Address reviews
* do not remove storPool snapshots
* add support for downloading diff snaps
* Add multiple zones support
* make copied snapshots have normal names
* address reviews
* Fix in progress
* continue fix
* Fix bulk delete
* change log to trace
* Start fix on multiple secondary storages for a single zone
* Fix multiple secondary storages for a single zone
* Fix tests
* fix log
* remove bitmaps when deleting snapshots
* minor fixes
* update sql to new file
* Fix merge issues
* Create new snap chain when changing configuration
* add verification
* Fix snapshot operation selector
* fix bitmap removal
* fix chain on different storages
* address reviews
* fix small issue
* fix test
---------
Co-authored-by: João Jandre <joao@scclouds.com.br>
* Update last agents during ms maintenance, and some code improvements
* Send 503 (Service Unavailable) response status when maintenance or shutdown is initiated
[Any load balancer in the clustered environment can avoid routing requests to this MS node]
* Migrate systemvm agents before routing host agents, and some code improvements
* Added events for ms maintenance and shutdown operations
* Added the following ms maintenance and shutdown improvements
- block new agent connections during prepare for maintenance of ms
- maintain avoids ms list
- propagate updated management servers list and lb algorithm in host and indirect.agent.lb.algorithm settings respectively, to systemvm (non-routing) agents
- updated setup ms list and migrate agent connections to executor service
- migrate agent connection through executor, and send the answer to the ms host that initiated the migration
- re-initialize ssl handshake executor if it is shutdown
- don't allow prepare for maintenance or shutdown when other management server nodes are in preparing states
- don't allow trigger shutdown when management server is up and other management server nodes are in preparing states
- stop agent connections monitor on ms maintenance
- update avoid ms list in ready command
- updated connected host from the client connection
- update last agents in ms metrics from the database
- updated some agent config descriptions
- update last management server in the hosts during shutdown
- added agents and lastagents in management server response
- updated management server maintenance & shutdown unit tests
- some code improvements
* refactored code / addressed comments
* removed shutdown testcase (maybe, calling System.exit)
* Revert "removed shutdown testcase (maybe, calling System.exit)"
This reverts commit e14b0717152ef6c8be102d61c80f42803a53172e.
* avoid system.exit during shutdown test
* code improvements
* testcase fix
* Fix cutoff time in agent connections monitor thread
* api,agent,server,engine-schema: scalability improvements
Following changes and improvements have been added:
- Improvements in handling of PingRoutingCommand
1. Added global config - `vm.sync.power.state.transitioning`, default value: true, to control syncing of power states for transitioning VMs. This can be set to false to prevent computation of transitioning state VMs.
2. Improved VirtualMachinePowerStateSync to allow power state sync for host VMs in a batch
3. Optimized scanning stalled VMs
- Added option to set worker threads for capacity calculation using config - `capacity.calculate.workers`
- Added caching framework based on Caffeine in-memory caching library, https://github.com/ben-manes/caffeine
- Added caching for account/use role API access with expiration after write can be configured using config - `dynamic.apichecker.cache.period`. If set to zero then there will be no caching. Default is 0.
- Added caching for account/use role API access with expiration after write set to 60 seconds.
- Added caching for some recurring DB retrievals
1. CapacityManager - listing service offerings - beneficial in host capacity calculation
2. LibvirtServerDiscoverer existing host for the cluster - beneficial for host joins
3. DownloadListener - hypervisors for zone - beneficial for host joins
5. VirtualMachineManagerImpl - VMs in progress- beneficial for processing stalled VMs during PingRoutingCommands
- Optimized MS list retrieval for agent connect
- Optimize finding ready systemvm template for zone
- Database retrieval optimisations - fix and refactor for cases where only IDs or counts are used mainly for hosts and other infra entities. Also similar cases for VMs and other entities related to host concerning background tasks
- Changes in agent-agentmanager connection with NIO client-server classes
1. Optimized the use of the executor service
2. Refactore Agent class to better handle connections.
3. Do SSL handshakes within worker threads
5. Added global configs to control the behaviour depending on the infra. SSL handshake could be a bottleneck during agent connections. Configs - `agent.ssl.handshake.min.workers` and `agent.ssl.handshake.max.workers` can be used to control number of new connections management server handles at a time. `agent.ssl.handshake.timeout` can be used to set number of seconds after which SSL handshake times out at MS end.
6. On agent side backoff and sslhandshake timeout can be controlled by agent properties. `backoff.seconds` and `ssl.handshake.timeout` properties can be used.
- Improvements in StatsCollection - minimize DB retrievals.
- Improvements in DeploymentPlanner allow for the retrieval of only desired host fields and fewer retrievals.
- Improvements in hosts connection for a storage pool. Added config - `storage.pool.host.connect.workers` to control the number of worker threads that can be used to connect hosts to a storage pool. Worker thread approach is followed currently only for NFS and ScaleIO pools.
- Minor improvements in resource limit calculations wrt DB retrievals
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
Co-authored-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
Co-authored-by: Rohit Yadav <rohit.yadav@shapeblue.com>
* test1, domaindetails, capacitymanager fix
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* test2 - agent tests
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* capacitymanagertest fix
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* change
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* fix missing changes
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* address comments
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* revert marvin/setup.py
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* fix indent
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* use space in sql
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* address duplicate
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* update host logs
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* revert e36c6a5d07
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* fix npe in capacity calculation
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* move schema changes to 4.20.1 upgrade
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* build fix
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* address comments
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* fix build
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* add some more tests
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* checkstyle fix
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* remove unnecessary mocks
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* build fix
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* replace statics
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* engine/orchestration,utils: limit number of concurrent new agent
connections
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* refactor - remove unused
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* unregister closed connections, monitor & cleanup
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* add check for outdated vm filter in power sync
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
* agent: synchronize sendRequest wait
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
---------
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
Co-authored-by: Rohit Yadav <rohit.yadav@shapeblue.com>
* Support for Management Server Maintenance
- New APIs: prepareForMaintenance and cancelMaintenance, with required parameter - managementserverid.
- New management server states for maintenance: PreparingForMaintenance, Maintenance.
- listHosts API with optional parameter – managementserverid, to list the hosts connected to the management server.
- Support management server maintenance when more than one active management servers available.
- Triggers transfer agents to other available management servers for maintenance, new agent command MigrateAgentConnectionCommand to initiate transfer of indirect agents.
- New global config 'management.server.maintenance.timeout', to set the timeout (in mins) for the management server maintenance window, default: 60 mins.
- UI changes: Prepare and Cancel Maintenance in Management Server section, Connected Agents tab, New fields for hosts and management servers.
* Updated pending jobs check timer task with ScheduledExecutorService
* keep maintenance state on trigger shutdown call when ms is in maintenance
* add pending jobs count to ms response
* during ms heartbeat, update state to up only when it's down
* allow vm work jobs of async job created before prepare for maintenance
* Revert "keep maintenance state on trigger shutdown call when ms is in maintenance"
This reverts commit 607e13364679eac897f4d146bb3325ea7a61ba17.
* skip maintenance test when multiple management servers are not available, and not configured in host setting for kvm
For HA work items that are created for host state change, checks must be
done when execution is called in a new management server session.
A new column, reason, has been added in cloud.op_ha_work table to track
the reason for HA work.
When HighAvailabilityManager starts it finds and puts all pending HA
work items in Investigating state. During execution of the HA work if it
is found in investigating state, checks are done to verify if the work
is still valid. If the jobs is found to be invalid it is cancelled.
Signed-off-by: Abhishek Kumar <abhishek.mrt22@gmail.com>
