Log in
Skip to sidebar
Skip to main content
Linked Applications
Loading…
Apache Software Foundation
Spaces
Hit enter to search
Help
Online Help
Keyboard Shortcuts
Feed Builder
What’s new
What’s new
Available Gadgets
About Confluence
Log in
Delete comments
Help
Delete all 0 comment(s) from this page
Delete selected comments
can you explain the use of the term backpressure here? It implies network traff...
Daan
|
Feb 23, 2017
back pressure created by several polling/activity/checks etc. on infra (network...
Rohit Yadav
|
Feb 27, 2017
no mixed zones!
Daan
|
Feb 23, 2017
Yes, the HA provider is configured per host basis. In a zone, with several clus...
Rohit Yadav
|
Feb 27, 2017
can you define it or specify where elagibility is defined (i.e. on a per provid...
Daan
|
Feb 23, 2017
Eligibility is defined by the HA provider driver, as a boolean method. The logi...
Rohit Yadav
|
Feb 27, 2017
where and how? reasonable defaults? configurable?
Daan
|
Feb 23, 2017
This is configurable both at global/cluster level. Reasonable defaults are base...
Rohit Yadav
|
Feb 27, 2017
can it define different types of checks, with different numbers? or is there on...
Daan
|
Feb 23, 2017
At the framework level, it simply understands and performs the number of activi...
Rohit Yadav
|
Feb 27, 2017
Should we have an extra state that indicate moving in and out of HA? i.e. a hos...
Daan
|
Feb 23, 2017
The disabled/ineligible state sort of tell us when a resource is moving in/out ...
Rohit Yadav
|
Feb 27, 2017
IPMI is only for hardware but this framework can be used for other resouces as ...
Daan
|
Feb 23, 2017
Yes. As an example, for the default HA provider for KVM we're shipping with the...
Rohit Yadav
|
Feb 27, 2017
So the initial state obviously is unknown but assumed disabled. Also they are t...
Daan
|
Feb 24, 2017
The New/Unknown state is not useful, otherwise the resource may be stuck indefi...
Rohit Yadav
|
Feb 27, 2017
is this differnet from when it has succeeded? i.e. other timeout or more extens...
Daan
|
Feb 24, 2017
For example, if in a round 10 checks can be performed – it will continue to do ...
Rohit Yadav
|
Feb 27, 2017
Do you mean the action resulting from the time out would be the same as a failu...
Daan
|
Feb 24, 2017
The activity check tasks are bounded by a timeout, on timeout it will raise an ...
Rohit Yadav
|
Feb 27, 2017
this should lead to scaling out to an extra ms or disabling HA altogether. A ha...
Daan
|
Feb 24, 2017
Just for this reason, we're saying that all the tasks/queues are bounded (i.e. ...
Rohit Yadav
|
Feb 27, 2017
and allowed to completed?
Daan
|
Feb 24, 2017
During FSM transition, from/to state are provided is they don't match exactly t...
Rohit Yadav
|
Feb 27, 2017
you mean maximum
Daan
|
Feb 24, 2017
I can change the wording, this is the amount of time we wait at least before at...
Rohit Yadav
|
Feb 27, 2017
you mean maximum
Daan
|
Feb 24, 2017
same as above
Rohit Yadav
|
Feb 27, 2017
operationHAForResource seems better as that implies the framework can later be ...
Daan
|
Feb 24, 2017
these are not generic APIs, but HA specific APIs only for hosts. In future if w...
Rohit Yadav
|
Feb 27, 2017
&& isHealthy == true
Daan
|
Feb 24, 2017
An eligible host can be unhealthy. Therefore, the check does not include health...
Rohit Yadav
|
Feb 27, 2017
according to the diagram this is DISABLED
Daan
|
Feb 24, 2017
You're right this can be removed.
Rohit Yadav
|
Feb 27, 2017
this one is DISABLED in the graph as well. I like the state table better;)
Daan
|
Feb 24, 2017
Err typo, this is RECOVERED.
Rohit Yadav
|
Feb 27, 2017
what happens if the manament server goes to or over the limit? what is the use ...
Daan
|
Feb 24, 2017
All the ephemeral queues are bounded. In case they are all full, HA operation s...
Rohit Yadav
|
Feb 27, 2017
recovering a recovered machine should be idempotent. the result should always b...
Daan
|
Feb 24, 2017
They are not idempotent from the resource perspective. By doing health/activity...
Rohit Yadav
|
Feb 27, 2017
Cancel
{"serverDuration": 94, "requestCorrelationId": "4c315dd78debcfca"}