Due to a regression (introduced by [1]), XenCenter will not run delayed actions in the correct host-order at all times. This can cause the master to be not the first host to be rebooted what may result in connections being lost to slaves should the API change between restarts/updates in a way that it would not be compatible to the previous version. This would make the wizard fail and users would have to restart their master host manually to make it connect to its slaves again.
This commit fixes the order of the hosts when executing the delayed actions.
[1]: 15827512e1
Signed-off-by: Gabor Apati-Nagy <gabor.apati-nagy@citrix.com>
In rare cases, the GetMaster could return null, causing the update progress to fail with "Pool_patch or Pool_update not found". (when the Cache hasn't been populated by the execution gets here)
This commit drops GetMaster and uses a pre-saved masterUuid instead of waiting for the Cache what would be the other way to solve this issue.
Signed-off-by: Gabor Apati-Nagy <gabor.apati-nagy@citrix.com>
Reset the PrecheckPage.PoolUpdate property on leaving the SelectPatch page, because it needs to be null for the Automated Updates mode and for the single update mode it will be updated on leaving the Upload page.
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
Perform all the prechecks on applicable hosts only (i.e. the hosts that don't have the update applied already), with the following exceptions:
- The HA check, which is performed on the pool master
- The server-side precheck (PatchPrecheckCheck) which will still be performed on all servers in the pool and will show the warning if the update has already been applied on some servers
This commit also adds an override to the HAOffCheck.SuccessfulCheckDescription to show the pool name instead of the master's if the check is successful.
Also, in the PatchPrecheckCheck, I had moved the code that checks if the patch is already applied to the very beginning (before the host liveness check), so it will return the warning that the server will be skipped even if the server is not reachable
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
Also rearranged some code in PatchingWizard class, to keep all property assignments together for each wizard page
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
Introduced the WizardMode enum and the WizardMode properties in several wizard pages (replacing the IsInAutomatedUpdates boolean property)
When in NewVersion mode, we show the same wizard pages as the AutomatedUpdates mode and we build the update sequence based on the selected update alert (which is an update to a new version)
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
In the Updates wizard:
New versions that are available as updates will be listed in the "Download Updates from Citrix" section, before any other updates.
If XenCenter is not the latest version, these rows will be greyed out, with a tooltip.
Signed-off-by: Gabor Apati-Nagy <gabor.apati-nagy@citrix.com>
- On the Updates tab: for new versions, check that the version of XenCenter is at least the latest. If not, give an additional option to download a newer XenCenter (for both updates and upgrades), and not offer "Download and Install" (for updates).
- In the Patching wizard: if the update is a new version, we force the user to first update their XenCenter to the new version
- In the RPU wizard: if the server version is not the latest and their XenCenter version is not the latest from updates.xml, we assume that they want to upgrade to the latest XenServer version and we force the user to first update their XenCenter to the latest version.
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
When a TargetInvocationException is raised, its InnerException property holds the underlying exception, which is what we should display.
Also ellipsise the action description (which could be an error thrown by the upload action) that is displayed on the Upload page.
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
This commit fixes a bug in the Update Wizard that causes XenCenter to not restart Toolstack when it should in the following case: in Automated Mode, when the update sequence for a host has an update that contains live-patch with restartHost guidance (and guidance-mandatory=false) followed by an update with Toolstack restart guidance.
Signed-off-by: Gabor Apati-Nagy <gabor.apati-nagy@citrix.com>
Also changed ApplyPatchAction and ApplyUpdateAction to install a single update to a single host, so that the appropriate action title can be displayed if an action fails. And replaced 2 hardcoded strings with resource strings.
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
Disabled the ability to select unreachable hosts and in addition
added a try/catch block to hide error message resulting from trying
to clean up an unreachable host.
Signed-off-by: Letsibogo Ramadi <letsibogo.ramadi@citrix.com>
This change does not apply to supplemental packs, as we should be able to install supp packs to individual hosts in a pool.
Also, if the wizard is installing an update from disk and we cannot match it in updates.xml, then we will continue to allow host selection, because we can't tell whether it is a supp pack or a hotfix.
Plus fixing the bug where the "Select All" button is enabled even when all the selectable servers/pools are selected
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
- Added a new method in XenTabPage which the derived classes can implement to select a default control after the page is loaded.
- Added a new wizard test that runs through the wizard by pressing the Enter key and checks if it has landed on the right page.
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
If batch updating fails in a pool, this code will do the clean up by
running the appropriate (next in the queue) RemoveUpdateFilesFromMaster action. In addition to Pool_update.clean(), Pool_update.destroy() will be called when needed.
Signed-off-by: Gabor Apati-Nagy <gabor.apati-nagy@citrix.com>
The Pool_update does not always covers all the VDIs XenCenter creates. These orphaned ones have to be deleted as well.
Signed-off-by: Gabor Apati-Nagy <gabor.apati-nagy@citrix.com>
For ISO updates, when disabling hosts, use the filename matching logic for pre-Ely hosts as well.
Signed-off-by: Gabor Apati-Nagy <gabor.apati-nagy@citrix.com>
Modify the cleanup pool update action to also call update destroy (removing the update from xapi db) when the update has not been applied to any hosts (ie we cancelled the wizard before completion).
This happened because the LivePatchCodesByHost was null when the prechecks were generated for an update, so the live patch codes returned by the precheck were ignored.
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
Due to the tooltip container introduced for the automatic radio button, winforms didn't treat these radio buttons properly - we needed code to make them mutually exclusive, tab didn't behave properly and keyboard arrow navigation between them didn't work. Therefore I have removed the tooltip container (so our radio buttons behave properly without any extra work) and we instead use a winforms tooltip shown when the mouse is over the automatic radio button. Note that since the control is disabled when we want a tooltip, neither hover events nor attaching a winforms tooltip in the usual way will work. Therefore I use a mousemove event on the table, which displays the tooltip if and only if we are over the disabled control. Also adjusted the tab indexes to make sure tab works correctly across the whole page.
Signed-off-by: Callum McIntyre <callumiandavid.mcintyre@citrix.com>
I've looked at how this was implemented in Dundee, and there we used an absolute height on the row containing the manual checkbox. I've done the same here, changing row 3 from autosize to absoliute 47px. I've also removed the empty row that was between the two radio buttons. This is a Winforms problem, without workarounds an autosize radio button in an autosize container (say the table row) will report its height as one row of text, no matter whether it needs more. This seems to be the only place we use a long radio label (as opposed to a separate label underneath the radio), so I doubt it's worth doing anything more complex than this. I've tested it with the CN string used in the ticket, and it uses less space than the English translation. The string also hasn't changed from Dundee.
Signed-off-by: Callum McIntyre <callumiandavid.mcintyre@citrix.com>
- added a new precheck, called "Live patching status" which checks is a reboot is needed for each host. If reboot is needed, the check will display a warning, otherwise it will show a green tickbox, saying that the server does not need to be rebooted, and if live patching is used, this information will be included as well.
Signed-off-by: Mihaela Stoica <mihaela.stoica@citrix.com>
