Skip to content

Fixing issue #57 - updating deployment manifests to v3.0.1#58

Open
marcelofernandez wants to merge 28 commits into
wunderio:masterfrom
ceresimaging:master
Open

Fixing issue #57 - updating deployment manifests to v3.0.1#58
marcelofernandez wants to merge 28 commits into
wunderio:masterfrom
ceresimaging:master

Conversation

@marcelofernandez
Copy link
Copy Markdown

No description provided.

marcelofernandez and others added 28 commits April 2, 2025 09:26
…rpart, using deterministic port generation from pod UUID
… UUID and improve service creation for rclone RC server. Linting.
Masking the hash value using '& 0xFFFFFFFF' to ensure it stays within a 32-bit range.

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
…ort handling, reusing previous getFreePort() functionality on cases where predicted port isn't available.
Expose rclone rc api to the cluster internal network
The rclone process in this pod serves every FUSE mount on the node. If
kubelet evicts the pod under node-pressure (for example when the image
filesystem crosses the eviction threshold), the rclone process dies and
every mount it was serving goes "Transport endpoint is not connected"
in every consumer pod. A new nodeplugin pod does not adopt the orphaned
mounts -- only recreating the consumer pods heals them.

Setting priorityClassName: system-node-critical keeps kubelet from
picking this DaemonSet as an eviction victim, so the rclone driver
survives disk / memory pressure instead of taking every FUSE mount on
the node down with it.
Mark nodeplugin DaemonSet as system-node-critical
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant