cloud/shepherd/equinix: implement recoverer
This implements basic recovery functionality for 'stuck' agents. The
shepherd will notice machines with a agent that either never sent a
heartbeat, or stopped sending heartbeats, and will remove their agent
started tags and reboot the machine. Then, the main agent start logic
should kick in again.
More complex recovery flows can be implemented later, this will do for
now.
Change-Id: I2c1b0d0465e4e302cdecce950a041581c2dc8548
Reviewed-on: https://review.monogon.dev/c/monogon/+/1560
Tested-by: Jenkins CI
Reviewed-by: Tim Windelschmidt <tim@monogon.tech>
diff --git a/cloud/shepherd/equinix/wrapngo/wrapn.go b/cloud/shepherd/equinix/wrapngo/wrapn.go
index 400556b..dc54340 100644
--- a/cloud/shepherd/equinix/wrapngo/wrapn.go
+++ b/cloud/shepherd/equinix/wrapngo/wrapn.go
@@ -113,6 +113,7 @@
// package comment for information on this method's behavior and returned error
// values.
UpdateSSHKey(ctx context.Context, kid string, req *packngo.SSHKeyUpdateRequest) (*packngo.SSHKey, error)
+ RebootDevice(ctx context.Context, did string) error
Close()
}
@@ -314,3 +315,11 @@
return k, nil
})
}
+
+func (e *client) RebootDevice(ctx context.Context, did string) error {
+ _, err := wrap(ctx, e, func(cl *packngo.Client) (struct{}, error) {
+ _, err := cl.Devices.Reboot(did)
+ return struct{}{}, err
+ })
+ return err
+}