-
Notifications
You must be signed in to change notification settings - Fork 70
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Update Reboot proto #99
base: main
Are you sure you want to change the base?
Conversation
@marcushines to review. |
Pull Request Test Coverage Report for Build 3160674502
💛 - Coveralls |
This will probably impact the following merged PRs: |
string reason = 4; // Reason for reboot. | ||
uint32 count = 5; // Number of reboots since active. | ||
} | ||
repeated Status statuses = 6; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
#not sure the logics correctly - if send a statusrequest that is empty what is the expected return? a statuses with a len of 1?
what if there is no reboot in process?
does the system return a single status message saying no reboot or empty statuses
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
should the Status have the path that is been queried as part of the response?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
-
ah, good catch. I copied and pasted the fields too fast and forgot to add the subcomponent paths. I just added it.
-
Reading the existing fields (e.g.
count, active
), seems like the original idea is to return reboot status (i.e. active=false) even no active reboot or planned reboot on the sub-component? The original idea makes sense to me.RebootStatusRequest
reports all reboots activities which covers the history and planning. Therefore, to align with the original idea I think we can clarify here saying that if sub-component not set, return status on all sub-components. Thoughts?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@xw-g The last entry of Reboot Status is covered in the openconfig-platform: last-reboot-reason.
leaf last-reboot-reason {
type identityref {
base oc-platform-types:COMPONENT_REBOOT_REASON;
}
description
"This reports the reason of the last reboot of the component.";
}
Should the RebootStatusRequest be applicable to only current/planned reboot requests?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
i would return error here if unset - as the total number of components may be large
@@ -108,7 +108,8 @@ message RebootRequest { | |||
uint64 delay = 2; | |||
// Informational reason for the reboot. | |||
string message = 3; | |||
// Optional sub-components to reboot. | |||
// Optional, sub-components to reboot. | |||
// If it's not set, reboot the whole network device. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
hmm i feel like we should require the component to reboot since rebooting a chassis as the default seems like a rather large hammer to "default" to
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This PR does require the user/client to be careful with the big hammer.
I had the similar concern, but I also debated with myself that if the field is required, then we probably want to have a very standard way (across all the vendors) to say this Foo
component/sub-component means the whole device for all vendors? Is it safe to use /components/component/name (identity=CHASSIS)
here? @dplore
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should the subcomponent
be renamed to component
? What is the logic behind naming it a subcomponent
? In openconfig-platform we have component -> subcomponent
relationship. If we assume that reboot works on paths defined in openconfig-platform, then maybe using a subcomponent name is not entirely logical.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes component makes sense - and if not set return not found or invalid parameter
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
then I suppose to reboot the chassis users should use the /components/component[name=chassis]
or something similar to indicate the the control plane modules should reboot?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We should require a single component. If the component is not specified, there should be an error.
Also agreed that we should explicitly clarify that reboot of /components/component[name=chassis] indicates a full system reboot.
types.Path subcomponent = | ||
1; // sub-component that the reboot status is for. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Unnecessary newline?
Hi all, If gNOI server runs on CPM_A, then what is expected of a system when you reboot this CPM or the whole chassis? Should the response be sent back before the component goes for a reboot? or should it just go for a reboot which results in a TCP session to break? /cc @marcushines |
If the concern is about accidentally rebooting a chassis, as opposed to a linecard, then perhaps the simplest thing to do would be to split these into two separate reboot RPC operations? E.g., RebootEntireDevice and RebootSubComponent? I'm also not sure that I see a great reason to allow multiple components to be rebooted in a single RPC, generally rebooting stuff should be pretty rare and not performance critical. |
// Optional, sub-components to reboot. | ||
// If it's not set, reboot the whole network device. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
// Optional, sub-components to reboot. | |
// If it's not set, reboot the whole network device. | |
// This string must contain the /components/component/name to reboot | |
// If not set, INVALID_ARGUMENT should be returned. | |
// If the request is accepted by the device, a gRPC OK should be returned. | |
// Reboot of /components/component[name=chassis] indicates a full system reboot. | |
// In the case of a full device reboot, a gRPC OK should still be returned and the gRPC | |
// session closed. This indicates to the client that the request was accepted and the | |
// reboot will proceed. | |
// |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
// In the case of a full device reboot without delay, a gRPC OK should still be returned and the gRPC
// session closed. This indicates to the client that the request was accepted and the
// reboot will proceed. With delay, gRPC OK will be returned
uint32 count = 5 [deprecated = true]; // Number of reboots since active. | ||
|
||
message Status { | ||
types.Path subcomponent = |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we use component instead of sub-component? A flat list of components should do the job. Any use cases for hierarchy of components can be determined via gNMI.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Component should be mandatory. The component tree is very large on most devices.
Internal bug b/245550570. |
@dplore I'm curious, is there a plan to move this PR forward? Been stale for almost a year. |
It's been pushed out for some time. I have queued it in my sprint plan, but currently is about 3 weeks out for review/refactoring. |
Propose to:
RebootStatusRequest
supports a list of subcomponents, theRebootStatusResponse
should also have a way to identify status per subcomponent.Another option is to deprecate the list of subcomponents in all reboot related proto, and only allow the request to target a single subcomponent. But then it raises more questions, like: