* fix fbeauchamp comments
* feat(backups): add a journal
* fix cache writing on immutable remotes
* add typing
* feat(backups): shard the journal per UTC day, one entry per time
`readBackupJournal()` listed the whole `xo-backup-log/` and filtered the names
client-side, so the cost of a read grew with the whole history of the repository.
Store the entries as `/xo-backup-log/<YYYYMMDD>/<HHMMSS.mmmZ>-<rand>-<event>-<vmUuid>-<file>`
and list only the days which can hold entries newer than `since`: a read now costs one
listing of the journal root plus one listing per day it is missing — normally a single
day, at most two, since a reader rebuilds from scratch when its watermark crosses a UTC
day. Purging old entries also becomes a directory removal instead of a per-object sweep.
The day is not repeated in the entry name, so `since` bounds the time of its own day
only; every later day is taken whole.
* fix pipeline
---------
Co-authored-by: All-Ki <killian.allegrain@vates.tech>
* feat(qcow2): improve generation speed
the idea is to build a bit index of the clusters to store them efficiently instead
of going through all the possible blocks 3 times at different phases of the generation.
The memory consumption is 2MB/TB of disk
this index generation can still be quite long , so we use setImmediate from
time to time to le the rest of the nodejs event loop run
also make this index phase abortable
* fix(xo-server): keep localhost proxy targets on wildcard binds
Since 82f047ef9 (#9572), [http.proxies] targets pointing at localhost are
rewritten to the hostname of the HTTPS listen config. With the default XOA
config (hostname = '0.0.0.0') this produced targets such as
http://0.0.0.0:9004/metrics and, worse, routed these internal requests
through the proxy-aware HTTP agent: with an httpProxy configured,
/openmetrics scrapes left the appliance for the corporate proxy and were
denied there.
Only rewrite when the bind address is a specific one (the #9500 case, where
localhost is not reachable) and pick the agent from the original target
locality, so requests to the machine itself never go through the HTTP proxy.
Seen in support ticket 7763112.
* chore(CHANGELOG.unreleased): entry for #10335
* chore(CHANGELOG.unreleased): clarify localhost targets always bypass the HTTP proxy
Opening the form in a new tab reloaded the whole app, so every object
was fetched again from scratch. Until that fetch finished, the form
could not be properly auto-filled. Navigating in the same tab reuses
the already loaded objects, so the form is populated right away.
docs(xo): fix wrong page order and restore missing links
After ve changed how the sidebar
was generated (from manual declarations to automatic generation)
Some pages went out of order and didn't
have the proper sidebar title.
This commit fixes that.
Also, the external links in the sidebar disappeared.
We can't add them in an auto-generated sidebar, so we circumvent
that by creating dedicated pages for external resources
and re-adding the links there.
Signed-off-by: Thomas Moraine <thomas.moraine@vates.tech>
We now generate the sidebar automatically (based on the documentation folder structure and _category.json files), instead of declaring the doc structure manually in sidebars.ts.
Simpler management
This makes the sidebar easier to understand and maintain. It is now managed the exact same way as the sidebars for the Vates VMS and XCP-ng documentations.
Clickable breadcrumbs
The autogenerated sidebar also creates category index pages, which makes breadcrumb categories clickable, as they already are in the Vates VMS and XCP-ng documentations.
Preview
Capture d'écran 2026-09-08 094848
Unlisted pages
Pages that should remain accessible but be hidden from the sidebar are now hidden using the unlisted: true metadata, instead of simply being omitted from sidebars.ts.
Internal links and redirection
Finally, the PR updates internal links and redirects to reflect the new folder structure and URLs.
* don't rearm after a fatal error before connection setting changed ( does not affect the normal route on non fatal error like newtork issue)
* serialize error correctly in database to ensure the "avoid a database write per auto-reconnect" corectly trigger
a lot of more cleanup to ensure that the connection is really dropped even if the host changed its id
* fix a mishnalding in the grace period before marking a pool disconnected . this will keep the ui in sync AND not redownload all the xapi object for a transient issue
* fix an error transmitted as raw, so the user was only seeing {}
[XO-2789]
* Extracts the backup migration logic into backup-archive and adds a cache guard for future performance improvements
* don't generate or read cache file on immutable backup repository
* clean cache file on immutable backup repository
* fix(backups): fetch failed
disable compresion between xoa on proxy on xoa side
* fix(xo-proxy): disable compression on the proxy
it was disable with hrp, replaced by #10038 since it was never
sending accept-encoding header
compression can be put back later on a route by route basis after
careful checks ( for example : file restore, backup listing )
* fix(backups): task label NBT->NBD
* fix: still force a list block only call for NBD + stream scenario , that was failing to transfer disk on fallback
* fix(openmetrics): keep the Prometheus secret across xo-server restarts
The `secret` property of `configurationSchema` used a random `default`. That
expression is re-evaluated every time the module is loaded, and xo-server never
persists the values ajv fills in from schema defaults, so each restart handed
the metrics endpoint a brand new bearer token and Prometheus started getting
401s.
Drop the default and generate the secret in `load()` instead, saving it through
`xo.configurePlugin()` so it survives a restart.
Introduced by #9323
See https://xcp-ng.org/forum/topic/12415
* chore(changelog): add openmetrics secret entry
* fix(xo-server-openmetrics): generate a 256-bit Prometheus secret
Buffer.from(getRandomValues(new Uint32Array(8))) coerces each 32-bit
element to a single byte, so the generated secret held 64 bits of
entropy instead of 256. Use randomBytes(32) instead.
Introduced by fe29bd0d4
* feat(lite): add create network and network bonded actions
* remove vlan required and info message
* feedback number one
* feat(lite): add create network and network bonded actions
* feat(lite): add action create internal network
* update changelog file
* fix rebase
* feedback number one
* remove as const in cancelRoute
* feedback number two
* feat(xo6): rebase detach
* add redirect when detach ok
* fix post rebase
* remove dev route
* returns from MR
* add check master host
* use linked translation
* rebase ok
* fix(backups): timeout was not used correctly for NBD transfers
* fix(nbd-client): evict a dead client in multi instead
actual code was retrying 5 times, each time with a 60s timeout
multi also have its own retry on conneciton
since the client used is determnistic for a block, any stalled client
will be hammerred for a lot of time before giving up
this PR evict a nbd client that is failing to read a block (with its
retry logic)
* fix(nbd-client/multi): better spread read accross client
the actual code derivate the clientId from the block index
we have no information on the block distribution, only hopes that
it's random enough
this PR introduce a mechanism to use a round robin to select
the next client
* fix: correct fall back to full with /without nbd on qcow2
