Hi Alex

We are looking at this. I will run a legacy test with your patches for 2144 and 
2145. I will also look a bit more into the consequences of this (SMF backwards 
compatibility, BC).
Note: A common way of handling BC is to add a configuration attribute to the 
general configuration IMM class that makes it possible to switch on and off a 
feature. However in this case you have implemented something that is specified 
in AIS so let's hope this is not needed.

Thanks
Lennart

> -----Original Message-----
> From: Alex Jones [mailto:[email protected]]
> Sent: den 28 februari 2017 01:42
> To: praveen malviya <[email protected]>; Lennart Lund
> <[email protected]>; Rafael Odzakow
> <[email protected]>; Neelakanta Reddy
> <[email protected]>
> Cc: [email protected]
> Subject: Re: [PATCH 1 of 1] amf: add support for restrictions to auto-repair
> [#2144]
> 
> Hi Guys,
> 
>   We need some consensus here to move tickets 2144 and 2145 forward.
> 
>   Here is my understanding of the AMF and SMF specs and current
> implementations regarding rolling upgrade with reboot. Please correct me
> if I am wrong. This mostly affects rollback and undo.
> 
>   I believe the SMF spec intends the behavior to be as follows for
> rolling upgrade with reboot (see 3.3.2.1 and 3.3.2.2):
> 
> (1) AMF lock of node
> (2) AMF lock-in of node
> (3) reboot node
> (3a) this reboot should not generate any DISABLED state transitions for
> any SUs as long as the components are properly terminated, otherwise the
> campaign goes into error detected state
> (4) AMF unlock-in of node
> (5) AMF unlock of node
> 
>   Can you point me to the place in the AMF spec where it says that a
> node must become disabled when it reboots? I can't seem to find it. The
> only thing I can find is 3.2.6.2:
> 
> "If a node is enabled and in the locked-instantiation administrative
> state when it leaves the cluster membership, the node stays enabled
> until it joins the cluster again."
> 
>   This says nothing about expected or unexpected leaving of the cluster.
> 
>   In short, I see two possible solutions:
> 
> (1) We need a way to reboot the node without AMF sending DISABLED
> operational states for SUs on that node, which it currently does even if
> the reboot is graceful. This is the preferred approach, as I believe it
> follows the specs. I have supplied a patch for this.
> 
> (2) Or, SMF could unset the suMaintenanceCampaign after it does the
> lock-in of the node before the reboot. The downside with this is that
> any errors in middleware SUs could not be detected.
> 
> Alex
> 
> On 02/27/2017 01:55 AM, praveen malviya wrote:
> > One more thing, as per spec AMF has to mark a node disabled when it
> > reboots.

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
Opensaf-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-devel

Reply via email to