Exosphere

🗳️ Suggestion: (At the risk of over-engineering this) using more information than just the message to identify the banner could help support cases where it perhaps should not remain hidden, for example: An increase in the level of severity of a ... [More] notice, a banner starts as info and escalates to danger while the banner maintainer doesn't update the content. An increase in the duration of a banner, the endsAt gets unexpectedly longer. One possible way to do this could use a hash function such as in the CopyableText widget (maybe combined with some leading chars from the message so that it's still easy to debug by inspection in local storage). [Less]

Overview Follow-up to #967 Banners are important messages from the system to each Exosphere user, however, once they have been seen, they can become clutter in the UI. In this change a new dismissal button is added to the banners so that someone ... [More] can choose to dismiss each one as they see fit. That dismissal is persisted in localStorage for continuity across browsing sessions. Once a particular banner is no longer being sent from the server, if it was dismissed, it will be removed from the dismissal list. This ensures that it will re-appear should it be brought back from the system. How to Test It can be helpful to rely on the browser’s ability to mock HTTP calls so you can force certain banner states from the server. When banners appear, look for the dismissal X on the right edge. Click on a dismissal and the banner should disappear. Reload the web page; restart the browser, etc… the dismissed banner should not reappear. Deactivate the HTTP mocking so that the real banners come from the server. The test banners should disappear as they normally would. Reactivate the HTTP mocking so that the fake banners reappear. The dismissed banners should now reappear, no-longer being dismissed because they were “recreated” on the server. Screenshots dismissable-banners-in-exosphere [Less]

Attaching Volumes to Shelved Instances This operation is more difficult & doesn't have a use-case in #1062 as burning as detaching an instance, so it's excluded from this MR. The operation is more difficult because it interacts with Better ... [More] names for volume mount points #831 which relies on server metadata for friendly naming. Server metadata can't be written to shelved_offloaded instances, Servers can transition from shelved to shelved_offloaded quite quickly, If the exoVolumes metadata isn't found when unshelving, the automount script's get_volume_name won't find a friendly name & the feature will break. There might be other ways to implement #831 (closed) or other ways to handle mounting, such as using a guest agent. I recommend exploring those as a separate item. [Less]

Overview Improves the loading state & display of server volumes in the tile on the Server Detail page. Removes the goToButton widget. Adds a "Volumes" link in the breadcrumbs of Attach and Mount Instructions pages. How to Test Check that ... [More] server volumes load correctly, links are followable & mount points are readable on the Server Detail page. Confirm references to the goToButton have been removed. When attaching a volume, note the "Volumes" link in the header breadcrumbs. Screenshots Server Volumes Tile Before After Breadcrumbs Before I found myself routinely wanting to visit my volumes list & having no easy way to navigate there. After [Less]

Overview How to Test Screenshots

Ideas: Check the Guacamole logs for clues See if there is anything on the Guacamole bug tracker: https://issues.apache.org/jira/browse/GUACAMOLE

Overview Fixes #1055 by excluding the (normally) transient reserved state when a volume is attached to a shelved instance. How to Test On the Project Overview & Volume List pages, check that network requests don't happen every 5 seconds when ... [More] you have at least 1 reserved instance in the project. Screenshots Before Screen_Recording_2025-04-15_at_10.31.51 After Screen_Recording_2025-04-15_at_10.30.22 Polling goes ahead as expected in other transient states. Screen_Recording_2025-04-15_at_10.39.00 [Less]

Chris Martin (973e9343) at 07 May 10:15 Merge branch 'fix/volume-polling' into 'master' ... and 1 more commit

Problem/Opportunity Statement Discovered during !1019, the Project Overview and Volume List switch to frequent volume polling if any volumes are in a transient state. The Reserved state is transient except when it applies to a shelved instance. ... [More] This results in perpetual frequent polling. Screen_Recording_2025-04-10_at_10.34.36 What would success / a fix look like? A check to determine whether volumes are reserved for a shelved instance would separate the transient from the long-lived statuses. Similar to !980, decreased polling would help with #612. [Less]