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/manager/fakequinix_test.go b/cloud/shepherd/equinix/manager/fakequinix_test.go
index 3970373..d9e5683 100644
--- a/cloud/shepherd/equinix/manager/fakequinix_test.go
+++ b/cloud/shepherd/equinix/manager/fakequinix_test.go
@@ -19,6 +19,7 @@
devices map[string]*packngo.Device
reservations map[string]*packngo.HardwareReservation
sshKeys map[string]*packngo.SSHKey
+ reboots map[string]int
}
// newFakequinix makes a fakequinix with a given fake project ID and number of
@@ -29,6 +30,7 @@
devices: make(map[string]*packngo.Device),
reservations: make(map[string]*packngo.HardwareReservation),
sshKeys: make(map[string]*packngo.SSHKey),
+ reboots: make(map[string]int),
}
for i := 0; i < numReservations; i++ {
@@ -168,5 +170,14 @@
return key, nil
}
+func (f *fakequinix) RebootDevice(_ context.Context, did string) error {
+ f.mu.Lock()
+ defer f.mu.Unlock()
+
+ f.reboots[did]++
+
+ return nil
+}
+
func (f *fakequinix) Close() {
}