Private/linl/sync #5461
Merged
Private/linl/sync #5461
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This is the initial commit for the update guidance improvement: 1. Update comments for re-purposing host.pending_guidances. 2. Add new pending-guidance fields: host.pending_guidances_recommended, host.pending_guidances_full, vm.pending_guidances_recommended and vm.pending_guidances_full. 3. Add definition for new guidance: RestartVM, reboot_host_on_kernel_livepatch_failure and reboot_host_on_xen_livepatch_failure 4. Add DB upgrade rule: replace reboot_host_on_livepatch_failure in "host.pending_guidances" with both reboot_host_on_kernel_livepatch_failure and reboot_host_on_xen_livepatch_failure in "host.pending_guidances_recommended” 5. Clear reboot_host_on_kernel_livepatch_failure and reboot_host_on_xen_livepatch_failure on reboot 6. CLI: expose those new added pending guidance fields Signed-off-by: Gang Ji <gang.ji@citrix.com>
As mandatory host guidance must be applied after applying updates, otherwise there may be e.g. VM failures, it is safe not to enable a host when its mandatory host guidance is pending: 1. do not enable host on xapi startup if there is pending mandatory guidance in the host. 2. raise error "host_pending_mandatory_guidances_not_empty" for API host.enable if there is pending mandatory guidance in the host. 3. not enable host when both on host startup and on only xapi startup. Signed-off-by: Gang Ji <gang.ji@citrix.com>
The guidance in new format in updateinfo.xml file deprecates the guidance in old format. For backwards compatibility support, both old and new guidance data will be in updateinfo.xml. With this commit, XAPI uses the guidance in new format only as the data in new format has been presented in updateinfo.xml. Signed-off-by: Ming Lu <ming.lu@cloud.com>
This commit updates the guidance evaluation logic and exposes the result of evaluation to the response of query on HTTP /updates endpoint. The response will be consumed by clients like XenCenter. The format of response is changed in this commit. Regarding this change, the backwards compatibility is not supported on this interface. Additionally, the evaluation is simplified by not considering the failed livapatches any more. Otherwise, the client like XenCenter has to introduce complicated UI design to populate the guidance change due to the livepatch failures. This may cause un-necessary carrying out of livepatch guidances. But the overall impact of carrying out these guidances is considered as ignorable. Signed-off-by: Ming Lu <ming.lu@cloud.com>
This commit updates the setting pending guidances logic. Since the host|VM.pending_guidances are re-purposed to be mandatory guidances, the setting logic is updated accordingly. Aslo fix-up setting pending mandatory guidances logic. The VMs' pending lists were missed. Meanwhile, the handling logic is changed to be easy for unit testing as well. Livepatch failures related guidance will not be in mandatory pending list. They will be handled in another pull request. Signed-off-by: Ming Lu <ming.lu@cloud.com>
This commit updates and adds unit tests for previous commits. Signed-off-by: Ming Lu <ming.lu@cloud.com>
Signed-off-by: Ming Lu <ming.lu@cloud.com>
…ge_from_master Merge master to update guidance improvement feature branch
Signed-off-by: Ming Lu <ming.lu@cloud.com>
This commit adds logic to set pending recommended and full guidance lists. What's more, the livepatch failures are merged with the pending livepatch failures. Livepatch failures are placed in pending recommended guidance list. Signed-off-by: Ming Lu <ming.lu@cloud.com>
Add unit tests for merge_livepatch_failures. Signed-off-by: Ming Lu <ming.lu@cloud.com>
In host.apply_updates, some safety checks are added in this commit: - disallow if there is any outstanding mandatory guidance on the host. - disallow if new mandatory guidance includes EvacuateHost or RebootHost and there are any running or paused VMs on the host. - automatically disable host (if not yet disabled). Signed-off-by: Ming Lu <ming.lu@cloud.com>
RestartVM is a task specified in update guidance. It has to be done on an updated host. Unlike other VM tasks in update guidance, a VM may escape from the tracking on individual host. E.g. a host evacuation will migrate a VM to aother host before host.apply_updates call setting the pending guidance on it. It's also not good to expose an API to set the pending guidance lists by clients. The solution in this commit is to set it for all running/paused/suspended VMs in the pool when the host.apply_updates is called for the coordinator host which is always the first one to be updated in a pool update. Signed-off-by: Ming Lu <ming.lu@cloud.com>
Signed-off-by: Ming Lu <ming.lu@cloud.com>
…46747 CP-46747: Expose 'title' field in updateinfo.xml to HTTP /updates
The RestartVM will be cleared when a VM's power state transits to Halted. So it is not good to add this update guidance action into shutdown VMs' pending guidance list, although it could be cleared in VM start also. Signed-off-by: Ming Lu <ming.lu@cloud.com>
…387034 CA-387034: RestartVM is added to pending guidances of shutdown VMs
Update document for error: "host_pending_mandatory_guidances_not_empty" as it is used not only when enabling host. Signed-off-by: Gang Ji <gang.ji@citrix.com>
Add field "last_update_hash" on host object to record the SHA256 checksum of updateinfo of the most recently applied update on the host. Signed-off-by: Gang Ji <gang.ji@citrix.com>
CP-43875: Record the repository hash on the host object when updating
The non-empty host pending mandatory guidance will result in the host blocked from being enabled or updated. This commit introduces a new API to clear the pending mandatory guidance in emergency cases. Otherwise, the host might be stuck in such a situation. Signed-off-by: Ming Lu <ming.lu@cloud.com>
A safety check in host.apply_updates will raise error if there is a mandatory EvacuateHost guidance but some VMs are running on the host. Meanwhile, EvacuateHost should not be added into pending list as it should be be carried out before host.apply_updates. Hence it needs to be filtered out. This commit fixes a bug that filering out the EvacuateHost before the safety check. It should be filtered out after the safety check. Signed-off-by: Ming Lu <ming.lu@cloud.com>
…45569 CP-45569: Add API Host.emergency_clear_mandatory_guidance
…up-safety-check Fixup: filter out EvacuateHost before safety check
…te channel When a user adds more update repositories as enabled ones, the pool.last_update_sync will be invalid as at the moment not all the enabled repositories have been synced at the time which is recorded in pool.last_update_sync. To avoid confusion, this commit resets the field to epoch so that the user can be notified to do a pool.sync_updates soon. Disabling an update repository will not make the pool.last_update_sync be invalid unless no enabled repositories after the disabling. Signed-off-by: Ming Lu <ming.lu@cloud.com>
There were 2 problems about host update guidance:
1. RestartToolstack guidance in pool member was cleared when there was
a xapi restart in pool coordinator.
This is because "consider_enabling_host_request" in pool member would be
called when there is a xapi restart in pool coordinator, as the db
connection to the coordinator will be re-established then.
2. CA-387695: RebootHost guidance in pool member got cleared when there
was a xapi restart in pool coordinator.
When xapi is started with a host (re)boot, its command line looks like:
/opt/xensource/bin/xapi -nowatchdog
-writereadyfile /var/run/xapi_startup.cookie
-writeinitcomplete /var/run/xapi_init_complete.cookie
-onsystemboot
then "!Xapi_globs.on_system_boot" will always be true.
Same as above, "consider_enabling_host_request" in pool member would be
called when there is a xapi restart in pool coordinator. And when
"!Xapi_globs.on_system_boot" is true for xapi in the pool member, the
RebootHost guidance in pool member would be cleared when there was a
xapi restart in pool coordinator.
These 2 problems are resolved by clearing host update guidance in
function "Dbsync_slave.update_env", which is called only during xapi
startup.
Signed-off-by: Gang Ji <gang.ji@citrix.com>
When a VM reboots, its qemu-dm-<domid> process restarts with its domid changed, but its RestartDeviceModel guidance was not cleared. This issue is resolved in this commit. When a VM starts, or reboots, xapi will be notified by xenopsd event to update its "last_start_time", RestartDeviceModel and RestartVM guidance is cleared there on VM start and VM reboot. When a VM is shutdown, "force_state_reset_keep_current_operations" will always be called to set the VM's state to "`Halted", RestartVM and RestartDeviceModel guidance is cleared there on VM shutdown. When a VM is suspended, "force_state_reset_keep_current_operations" will always to called to set the VM's state to "`Suspended", RestartDeviceModel guidance is cleared there on VM suspend. When a VM is migrated, "pool_migrate_complete" will always be called, RestartDeviceModel guidance is cleared there on VM migration. Below are the tested scenarios: VM running -> (force)halted result: RestartDeviceMode RestartVM cleared if any VM running -> paused result: Nothing VM running -> suspended result: RestartDeviceMode cleared if any VM halted -> running result: RestartDeviceMode RestartVM cleared if any VM paused -> running result: Nothing VM paused -> (force)halted result: RestartDeviceMode RestartVM cleared if any VM paused -> (force)reboot result: RestartDeviceMode RestartVM cleared if any VM suspend -> running result: Nothing VM suspend -> force halted result: RestartDeviceMode RestartVM cleared if any VM running -> (force)reboot result: RestartDeviceMode RestartVM cleared if any VM running -> take snapshot with no memory result: Nothing VM running -> take snapshot with memory result: RestartDeviceMode cleared if any VM running(with RestartDeviceMode and RestartVM guidance) -> take snapshot with no memory -> revert to snapshot with no memory result: both RestartDeviceMode RestartVM cleared VM running(with RestartDeviceMode and RestartVM guidance) -> take snapshot with memory -> add RestartDeviceMode guidance -> revert to snapshot with memory result: RestartDeviceMode cleared, RestartVM remains VM running(with RestartDeviceMode and RestartVM guidance) -> take snapshot with no memory -> clear RestartDeviceMode RestartVM guidance by reboot -> revert to snapshot with no memory result: both RestartDeviceMode and RestartVM cleared VM running(with RestartDeviceMode and RestartVM guidance) -> take snapshot with memory -> clear RestartDeviceMode RestartVM guidance by reboot -> revert to snapshot with memory result: only RestartDeviceMode cleared, RestartVM restored One side effect from this change: In case there is any RestartVM or RestartDeviceModel guidance on a halted VM, then a toolstack restart on pool master will clear those VM guidance from the halted VMs. However it should not be a problem as: 1. xapi will not add VM guidance on halted VMs 2. once a VM turns to halted state, the VM guidance will be cleared by xapi So there should not be any VM guidance on halted VMs. Signed-off-by: Gang Ji <gang.ji@citrix.com>
The checking task status logic in CLI server side is used by multiple interfaces. It's better to split it out to resolve the duplications. Signed-off-by: Ming Lu <ming.lu@cloud.com>
Signed-off-by: Ming Lu <ming.lu@cloud.com>
Signed-off-by: Ming Lu <ming.lu@cloud.com>
Signed-off-by: Ming Lu <ming.lu@cloud.com>
Signed-off-by: Ming Lu <ming.lu@cloud.com>
…45572 Add "xe host-show-updates" CLI and update "xe host-apply-updates"
Signed-off-by: Gang Ji <gang.ji@citrix.com>
…rom_master.2 Merge master into update guidance improvement feature branch
So that the existing XenCenter will refuse to connect to the new XAPI. This aims to force users to use the new XenCenter since the old XenCenter can't work with the new XAPI with update guidance improvement. Signed-off-by: Gang Ji <gang.ji@citrix.com>
…new xapi Add livepatch guidance DB upgrade rule: except reboot_host_on_livepatch_failure, move any guidance in "host.pending_guidances" into "host.pending_guidances_recommended". Signed-off-by: Gang Ji <gang.ji@citrix.com>
Signed-off-by: Ming Lu <ming.lu@cloud.com>
Signed-off-by: Ming Lu <ming.lu@cloud.com>
…ate-datamodel-lifecycle Update datamodel lifecycle for update guidance feature
The default value of 'flags' is '[`Session; `Async]'. Explicit value '[`Session]' caused no async support on this API method. The fix is to use the default value. Signed-off-by: Ming Lu <ming.lu@cloud.com>
…388699 CA-388699: No async support on VM.restart_device_models
Signed-off-by: Konstantina Chremmou <Konstantina.Chremmou@cloud.com>
[ASAP] CP-47509: Expose RequestHeaders and ResponseHeaders in C# SDK.
the data field is not part of meta. see the XML generator at ocaml/libs/xapi-rrd/lib/rrd_updates.ml line 124 The issue was introduced when serialization was changed to use yojson: xapi-project/xcp-rrd@048b0e3 Signed-off-by: Pau Ruiz Safont <pau.ruizsafont@cloud.com>
Signed-off-by: Rob Hoes <rob.hoes@citrix.com>
Signed-off-by: Rob Hoes <rob.hoes@citrix.com>
…ance-improvement Merge feature/update-guidance-improvement branch to master
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## feature/py3 #5461 +/- ##
============================================
Coverage 49.07% 49.07%
============================================
Files 18 18
Lines 2319 2319
============================================
Hits 1138 1138
Misses 1181 1181
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
gangj
approved these changes
Feb 20, 2024
There was a problem hiding this comment.
The last commit in the source branch https://github.com/liulinC/xen-api/tree/private/linl/sync is 34dabdf, it is just the current latest commit in master.
bernhardkaindl
approved these changes
Feb 20, 2024
pytype_reporter extracted 50 problem reports from pytype output. You can check the results of the job here |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Just routinely feature branch sync,
No any update on feature branch.