We've been running a CMG on ConfigMgr with token based auth, and when set up in version 2002, it was a single cloud (classic) VM and this has been working well. We did the MECM upgrade to version 2107 a couple of weeks ago and updated clients, all well there too.
However, from yesterday, I've noted that a low number of clients are connected via the CMG, normally this is 400-500 but only seeing around 100-110 at present, and even then a number of them seem to then go offline.
I did note in CloudMgr.log that this appears to happen every five minutes, a recent snippet as below:
Current CMG deployment package hash is nvuhjuOw53Mscz26rqPXpS8LHupH2asSYb2A+YbRBcM= SMS_CLOUD_SERVICES_MANAGER 22/09/2021 10:03:48 12160 (0x2F80)
Current VMSS CMG deployment package hash is 151sGQMFOWHmByoADw4+CUUh7ypg/ix4BIjb9jILIEg= SMS_CLOUD_SERVICES_MANAGER 22/09/2021 10:03:48 12160 (0x2F80)
CreateUpgradeTaskForPackageChange: Triggered to create upgrade task for service type CloudProxyService with current package hash nvuhjuOw53Mscz26rqPXpS8LHupH2asSYb2A+YbRBcM=. SMS_CLOUD_SERVICES_MANAGER 22/09/2021 10:03:48 12160 (0x2F80)
TaskWorker: Starting task: [CloudServicesTaskBuilder] SMS_CLOUD_SERVICES_MANAGER 22/09/2021 10:03:48 5684 (0x1634)
CloudServicesTaskBuilder: Starting. SMS_CLOUD_SERVICES_MANAGER 22/09/2021 10:03:48 10920 (0x2AA8)
TaskManager: 0 task(s) running, 0 task(s) waiting to start. SMS_CLOUD_SERVICES_MANAGER 22/09/2021 10:03:48 11016 (0x2B08)
TaskManager: Task [CloudServicesTaskBuilder] status is Running SMS_CLOUD_SERVICES_MANAGER 22/09/2021 10:03:48 11016 (0x2B08)
CloudServicesTaskBuilder: Stopping. SMS_CLOUD_SERVICES_MANAGER 22/09/2021 10:03:48 10920 (0x2AA8)
TaskManager: 1 task(s) running, 0 task(s) waiting to start. SMS_CLOUD_SERVICES_MANAGER 22/09/2021 10:03:48 11016 (0x2B08)
TaskManager: Task [CloudServicesTaskBuilder] status is RanToCompletion SMS_CLOUD_SERVICES_MANAGER 22/09/2021 10:03:48 11016 (0x2B08)
The CMG itself from the ConfigMgr console shows the status of Ready, and the connection analyzer when run shows that it has passed all of the six tests with a green tick. It's almost as if that task to upgrade the cloud proxy service exits out.
From the Azure side, when checking the operational logs, it appears to show a fail when changing config, and the JSON then shows that there's no change in config. However, the operational logs only appear to show once every two hours.
Is this constant upgrade attempting meaning that the cloud service isn't allowing periodic responses, hence kicking clients off, or is there something else I need to check?