If health checks are failing for all servers, make a second pass through the server list with health-checks ignored before returning failure
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
(cherry picked from commit ca39614d4e)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
It is concievable that users might take more than 60 seconds to deploy their own cloud-provider. Instead of exiting, we should wait forever, but with more logging to indicate what's being waited on.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
(cherry picked from commit ed23a2bb48)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Start shared informer caches when k3s-etcd controller wins leader election. Previously, these were only started when the main k3s apiserver controller won an election. If the leaders ended up going to different nodes, some informers wouldn't be started
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
(cherry picked from commit 3d14092f76)
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Prefer the address of the etcd member being joined, and seed the full address list immediately on startup.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
* Adds support for health-checking loadbalancer servers. If a
health-check fails when dialing, all existing connections to the
server will be closed.
* Wires up a remotedialer tunnel connectivity check as the health check
for supervisor/apiserver connections.
* Wires up a simple ping request to the supervisor port as the health
check for etcd connections.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Prevents joining nodes from being stuck with bad initial member list if there is a transient failure, or if they try to join themselves
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
* Set ServerNodeName in snapshot CLI setup
* Raise errer if ServerNodeName ends up empty some other way
* Fix status controller to use etcd node name annotation instead of prefix checking
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
If a full reconcile wins the race against sync of an individual snapshot resource, or someone intentionally deletes the configmap, the data map could be nil and cause a crash.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
While some implementations may support it, it appears that most don't,
and some may in fact return an error if it is requested.
We already stat the object to get the metadata anyway, so this was
unnecessary if harmless on most implementations.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Omit snapshot list configmap entries for snapshots without extra metadata; reduce log level of warnings about missing s3 metadata files.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
These fields are only necessary when saving snapshots to S3, and will block restoration if attempted
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Silences error message from lasso - this is a normal startup condition
when no snapshots exist so we shouldn't log nasty looking errors.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Removing this in 002e6c43ee regressed
control-plane-only nodes, as we rely on the etcd client to update its
endpoint list internally so that we can use it to sync the load-balancer
address list.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Write the extra metadata both locally and to S3. These files are placed such that they will not be used by older versions of K3s that do not make use of them.
Signed-off-by: Brad Davidson <brad.davidson@rancher.com>
Brad Davidson