Hi Minh,
In the second patch, assignment messages are buffered. These will be sent to
AMFD after headless state.
There are two broad cases (Discusses earliar also):
1)When comp completes assignments in headless state.
2)When atleast one calback is pending after headless.
In case1), AMFD can self trigger SG FSM as there will not be any assignment
message pending from AMFND. For this AMFD will have to rely on the SUSI FSM
states from AMFND as those will be updated.
In case 2), since assignment is pending to comp, after headless state comp will
respond for the callback. This will create non-headless type assignment event
that will trigger SG FSM at AMFD. So in this case, AMFD has to wait for event
from AMFND,
So in this way, buffereing may not be required.
Also in #1902 patch, recovery request is being buffered. If a comp faults after
responding for the callback, then AMFND will have to buffer assignment events
as well as recovery request. This needs to be given in the same order to AMFD.
Do you see any problem if things are done without buffering assignment response
at AMFND?
Also what I see is, patch 1 completely relies on SUSI FSM states that it gets
from IMM and not from the Si assignment status that it gets from AMFND. Can't
AMFD work with the states that it gets from AMFND when node is available? I
think for the failed-rebooted nodes only we wanted to have things read from
IMM.
Thanks,
Praveen
---
** [tickets:#1725] AMF: Recover transient SUSIs left over from headless**
**Status:** review
**Milestone:** 5.1.FC
**Created:** Wed Apr 06, 2016 07:16 AM UTC by Minh Hon Chau
**Last Updated:** Tue Aug 09, 2016 12:57 PM UTC
**Owner:** Minh Hon Chau
This ticket is more likely an enhancement that targets on how AMFD detect and
recover the transients SUSI left over from headless. There are three major
situations:
(1) - Cluster goes headless, su/node failover on any payloads can happen, or
any payloads can be hard rebooted/powered off by operator, then cluster recover
(2) - issue admin op on any AMF entities, cluster goes headless. During
headless, the middle HA assignments of whole admin op sequence between AMFND
and components could be:
(2.1) The assignment completes, component returns OK with csi callback,
then cluster recover
(2.2) The assignment is under going, then cluster recover. The assignment
afterward could complete, or csi callback returns FAILED_OPERATION or error can
also happen
At the time cluster recover, amfd has collected all assignments from all
amfnd(s). These assignments can be in assigned or assigning states whilst its
HA states do not conform its SG redundancy. Any of (1) (2.1) (2.2) can happen
in a combination, which means while issuing admin op (2), cluster go headless
and any kinds of failover (1) can happen during headless.
---
Sent from sourceforge.net because [email protected] is
subscribed to https://sourceforge.net/p/opensaf/tickets/
To unsubscribe from further messages, a project admin can change settings at
https://sourceforge.net/p/opensaf/admin/tickets/options. Or, if this is a
mailing list, you can unsubscribe from the mailing list.------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are
consuming the most bandwidth. Provides multi-vendor support for NetFlow,
J-Flow, sFlow and other flows. Make informed decisions using capacity
planning reports. http://sdm.link/zohodev2dev
_______________________________________________
Opensaf-tickets mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-tickets
