When the functionality for remote system management (scheduling tasks on client systems) is designed, could the following please be taken into account:
Current Satellite uses either regular rhnsd check-in or prompts clients to immediately check-in via osad. There seems to be no middle ground.
For sites with thousands of Katello clients, it could be ruinous to schedule tasks to operate on all clients immediately. It is acknowledged that some tasks will be required immediately, but thought should be given to allow a more relaxed schedule for other tasks.
Could the following scheduling scenarios please be considered for Katello (regardless of the date/time set for a task):
1) Select date/time for task to begin
2) Select scheduling method to be used for *this* task (Immediate/Regular)
Tasks scheduled on the Regular scheduling method would only be picked up on the once-every-4-hours rhnsd style check-in (configurable I know).
Tasks scheduled on the Immediate scheduling method would result in the clients being prompted to carry out tasks as soon as possible after the scheduled date/time.
I understand AMQP will possibly be used to arrange this. Not sure how that affects things.
Under the current Satellite regime, if clients are configured with Push-to-Client, all scheduled tasks are carried out via osad. There is no option to allow a task to be picked up only on a regular rhnsd check-in.
On small estates, this is not an issue.
When estates are scaled up to thousands or tens of thousands of clients, there are implications to scheduling tasks across all clients.
1) On a highly (or completely) virtualised environment, loading on the virtual hosts could rise to alarm levels if complex tasks are scheduled immediately across all clients
2) Load on the central Katello/Satellite system/smart-proxies could rise to unacceptable levels
It would be polite to the infrastructure if tasks could be jittered in a way that spreads the load.
Certain tasks may be required immediately, so any jittering should be user defineable on a task by task basis. i.e. jitter=0 for urgent tasks, jitter=4 hours for tasks that I'm more relaxed about.
Examples of such tasks:
Urgent: A failure of my single (bad design I know) NTP server prompts me to build a temporary NTP server. I'll want to push out a new config and restart NTP immediately.
Relaxed: IT Security require a report (RPM verify for example) from all my machines by tomorrow. I'd like to schedule the task to start outside business hours, but be jittered by 4 hours to ensure no spike load hits the virtual hosts.
Moving to Sat6 to be tracked there. Upstream bugs are moving to redmine.
Since this issue was entered in Red Hat Bugzilla, the release flag has been
set to ? to ensure that it is properly evaluated for this release.
Reset docs contact <> daobrien
Verified in Satellite 6.2 Beta Snap 15.2
This feature has been fully delivered in Satellite 6.2 in the form of the Remote Execution plugin.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.