The following is the scope and behaviour of this new feature that we have come
up with after initial analysis. Comments are welcome.
1)Node group class "SaAmfNodeGroup" will be enhanced by including attribute
"saAmfNGAdminState".
<attr>
<name>saAmfNGAdminState</name>
<type>SA_UINT32_T</type>
<category>SA_RUNTIME</category>
<flag>SA_PERSISTENT</flag>
<flag>SA_CACHED</flag>
<default-value>1</default-value>
</attr>
2)Only shutdown and unlock admin operation will be supported on Nodegroup.
3)Shutdown operation on Nodegroup will be supported only for Nored and
Nway_active
redudancy model.
4)Nodegroup admin state "saAmfNGAdminState" will show admin operation status.
5)When shutdown operation is initiated on the nodegroup, all the "application"
SUs (not middleware SUs) deployed on the nodes of that nodegroup will be
unassigned through quiescing state. This is as if each node is undergoing a
shutdown operation,but saAmfNodeAdminState will remain unlocked. Thus there
will be only graceful removal of assignment from SUs and not the termination of
SUs from the nodes during nodegroup shutdown operation. i.e. There is no
lock-in behaviour.
6) Shutdown operation on a nodegroup will be rejected in the following cases:
a)If any AMF entity is not stable on any node of nodegroup.
b)If two or more SUs of same SG are hosted on any node of the nodegroup.
c)If any node of the nodegroup is already under going any other admin
operation.
d)If any SU on any node of the nodegroup is undergoing any admin op.
7)As per AMF Pr doc, AMF does not support SI deps in NoRed Model and
Nway_active model. So
SI dependency will not be honoured during shutdown operation while assigning
quiescing state to SUs or removal of assignemet from SUs hosted on the nodes of
nodegroup.
Return values for the shutdown admin operation on nodegroup:
1)SA_AIS_OK- The function completed successfully.
2)SA_AIS_ERR_TIMEOUT - An implementation-dependent timeout occurred before the
call could complete. It is unspecified whether the call succeeded or whether it
did not.
3)SA_AIS_ERR_TRY_AGAIN- This will be returned in the following cases:
a)If any AMF entity is not stable on any node of nodegroup.
b)If any node of the nodegroup is already under going any other admin
operation.
c)If any SU on any node of the nodegroup is undergoing any admin op.
4)SA_AIS_ERR_NOT_SUPPORTED-If other than shutdown and unlock operation are
operated on nodegroup. Or if two or more SUs of same SG are hosted on any node
of nodegroup.
5)SA_AIS_ERR_BAD_OPERATION - If shutdown operation is initiated on a nodegroup
which is already in locked state.
6)SA_AIS_ERR_NO_OP - If another shutdown operation is initiated on the
nodegroup which is already in shutting-down state
because of previous shutdown operation.
Return values for the unlock admin operation on nodegroup:
1)SA_AIS_OK - The function completed successfully.
2)SA_AIS_ERR_TIMEOUT - An implementation-dependent timeout occurred before the
call could complete. It is unspecified whether the call succeeded or whether it
did not.
3)SA_AIS_ERR_TRY_AGAIN-If any AMF entity is not stable on any node of nodegroup.
4)SA_AIS_ERR_NO_OP - The invocation of this administrative operation has no
effect
on the current state of the logical entity, as it is already in unlocked state.
---
** [tickets:#1235] NodeGroup shutdown Admin op to support parallel admin
shutdown on multiple nodes**
**Status:** assigned
**Milestone:** future
**Created:** Wed Dec 10, 2014 12:25 PM UTC by Mathi Naickan
**Last Updated:** Tue Dec 23, 2014 11:12 AM UTC
**Owner:** Praveen
Today, it is not possible to initiate AMF node shutdown operation on multiple
nodes at the same time.
This ticket proposes (praveen's idea) to introduce a new AMF admin operation on
the nodegroup logical entity.
A cluter scale-down usecase might introduce limits on the total time spent in
down-sizing an AMF/application cluster. This enhancement shall look into
different ways by which atleast some amount of parallelism is induced/pushing
within AMF (wherever applicable) during the processing of this new 'node-group'
shutdown operation.
End of the day, the final timetake will be a mix of
1) amount of time spent in the checks inside AMF during processing of a
parallel admin_op.
+
2) how quicky applications respond to the csi callbacks.
Atleast to start with one redundancy model...
More later... Nagendra, Praveen would update on the details on the scenarios in
which AMF would still continue to do serial processing i.e. where existing
checks cannot be removed, etc....
(The bigger opportunity is also not to overlty complicate AMF by introducing
this adminop)
---
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.------------------------------------------------------------------------------
New Year. New Location. New Benefits. New Data Center in Ashburn, VA.
GigeNET is offering a free month of service with a new server in Ashburn.
Choose from 2 high performing configs, both with 100TB of bandwidth.
Higher redundancy.Lower latency.Increased capacity.Completely compliant.
http://p.sf.net/sfu/gigenet
_______________________________________________
Opensaf-tickets mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-tickets
