> -----Original Message-----
> From: praveen malviya [mailto:[email protected]]
> Sent: den 16 januari 2014 11:54
> To: Hans Feldt
> Cc: [email protected]
> Subject: Re: [devel] [PATCH 1 of 1] amfnd: escalate comp errors in quiesced
> HA state [#601]
>
> Hi,
> I am testing the patch and got one doubt.
> While removing the quiesced assignment, if component fault then also
> compoent failover is escalated.
> Do we require component-failover escalation at the the time of removal
> of assignments through quiesced state.?
Well not in reality since at that point in time there would exist a new active
assignment for some other component. But what should amfnd reassign to the
restarted component in this case?
Nothing I guess until it gets a standby assignment.
I prefer not to complicate it even more and handle all errors the same. Makes
it easier to maintain, understand and explain to others.
Thanks,
Hans
> Or component-failover escalation is required only when component faults
> when AMF is assigning quiesced state to it ?
>
>
> Thanks
> Praveen
>
>
> On 10-Jan-14 2:44 PM, Hans Feldt wrote:
> > osaf/services/saf/amf/amfnd/comp.cc | 23
> > +++++++++++++++++++++--
> > osaf/services/saf/amf/amfnd/err.cc | 15 +++++++++++----
> > osaf/services/saf/amf/amfnd/include/avnd_comp.h | 1 +
> > osaf/services/saf/amf/amfnd/su.cc | 1 +
> > osaf/services/saf/amf/amfnd/util.cc | 5 ++---
> > 5 files changed, 36 insertions(+), 9 deletions(-)
> >
> >
> > When a node or SU is locked, AMF assigns the QUIESCED HA state to ACTIVE
> > sa-aware components. This state is intended to transfer state to a peer
> > component assigned STANDBY. If the ACTIVE component fails in the QUIESCED
> > state, AMF will restart it and reassign it QUIESCED. This is an invalid
> > state
> > transition according to picture 3 in the B.04 AMF spec and makes no sense
> > since
> > the state is anyway gone (because the component failed).
> >
> > The problem was likely reintroduced when fixing #3083, before that special
> > handling for comp errors in QUIESCED state was there. However the solution
> > had
> > other problems and was removed. So in the current code there is no special
> > handling for errors in QUIESCED state which results in re-assignment of the
> > QUIESCED HA state after component restart.
> >
> > This patch changes the error escalation logic so that instead of component
> > restart, component failover is performed as recovery action.
> >
> > diff --git a/osaf/services/saf/amf/amfnd/comp.cc
> > b/osaf/services/saf/amf/amfnd/comp.cc
> > --- a/osaf/services/saf/amf/amfnd/comp.cc
> > +++ b/osaf/services/saf/amf/amfnd/comp.cc
> > @@ -32,9 +32,7 @@
> >
> > ******************************************************************************
> > */
> >
> > -#include <stdbool.h>
> > #include "avnd.h"
> > -#include <stdbool.h>
> > #include <immutil.h>
> >
> > /*** Static function declarations ***/
> > @@ -2766,3 +2764,24 @@ void avnd_comp_pres_state_set(AVND_COMP
> > m_AVND_SEND_CKPT_UPDT_ASYNC_UPDT(cb, comp, AVND_CKPT_COMP_PRES_STATE);
> > }
> >
> > +/**
> > + * Returns true if the HA state for any CSI assignment is
> > QUIESCED/QUIESCING
> > + * @param su
> > + */
> > +bool comp_has_quiesced_assignment(const AVND_COMP *comp)
> > +{
> > + const AVND_COMP_CSI_REC *csi;
> > +
> > + for (csi = m_AVND_CSI_REC_FROM_COMP_DLL_NODE_GET(
> > + m_NCS_DBLIST_FIND_FIRST(&comp->csi_list));
> > + csi != NULL;
> > + csi = m_AVND_CSI_REC_FROM_COMP_DLL_NODE_GET(
> > + m_NCS_DBLIST_FIND_NEXT(&csi->comp_dll_node))) {
> > +
> > + if ((csi->si->curr_state == SA_AMF_HA_QUIESCED) ||
> > + (csi->si->curr_state == SA_AMF_HA_QUIESCING))
> > + return true;
> > + }
> > +
> > + return false;
> > +}
> > diff --git a/osaf/services/saf/amf/amfnd/err.cc
> > b/osaf/services/saf/amf/amfnd/err.cc
> > --- a/osaf/services/saf/amf/amfnd/err.cc
> > +++ b/osaf/services/saf/amf/amfnd/err.cc
> > @@ -411,10 +411,17 @@ uint32_t avnd_err_escalate(AVND_CB *cb,
> > if (*io_esc_rcvr == SA_AMF_NO_RECOMMENDATION)
> > *io_esc_rcvr = comp->err_info.def_rec;
> >
> > - /* disallow comp-restart if it's disabled */
> > - if ((SA_AMF_COMPONENT_RESTART == *io_esc_rcvr) &&
> > m_AVND_COMP_IS_RESTART_DIS(comp) && (!su->is_ncs)) {
> > - LOG_NO("saAmfCompDisableRestart is true for
> > '%s'",comp->name.value);
> > - *io_esc_rcvr = SA_AMF_COMPONENT_FAILOVER;
> > + if (*io_esc_rcvr == SA_AMF_COMPONENT_RESTART) {
> > + if (m_AVND_COMP_IS_RESTART_DIS(comp) && (!su->is_ncs)) {
> > + LOG_NO("saAmfCompDisableRestart is true for
> > '%s'",comp->name.value);
> > + LOG_NO("recovery action 'comp restart' escalated to
> > 'comp failover'");
> > + *io_esc_rcvr = SA_AMF_COMPONENT_FAILOVER;
> > + } else if (comp_has_quiesced_assignment(comp) == true) {
> > + /* Cannot re-assign QUIESCED, escalate to failover */
> > + LOG_NO("component with QUIESCED/QUIESCING assignment
> > failed");
> > + LOG_NO("recovery action 'comp restart' escalated to
> > 'comp failover'");
> > + *io_esc_rcvr = SA_AMF_COMPONENT_FAILOVER;
> > + }
> > }
> >
> > if ((SA_AMF_COMPONENT_FAILOVER== *io_esc_rcvr) && (su->sufailover) &&
> > (!su->is_ncs)) {
> > diff --git a/osaf/services/saf/amf/amfnd/include/avnd_comp.h
> > b/osaf/services/saf/amf/amfnd/include/avnd_comp.h
> > --- a/osaf/services/saf/amf/amfnd/include/avnd_comp.h
> > +++ b/osaf/services/saf/amf/amfnd/include/avnd_comp.h
> > @@ -876,6 +876,7 @@ extern unsigned int avnd_comp_config_get
> > extern int avnd_comp_config_reinit(AVND_COMP *comp);
> > extern void avnd_comp_delete(AVND_COMP *comp);
> > extern void avnd_comp_pres_state_set(AVND_COMP *comp, SaAmfPresenceStateT
> > newstate);
> > +bool comp_has_quiesced_assignment(const AVND_COMP *comp);
> >
> > /**
> > * Initiate restart of a component.
> > diff --git a/osaf/services/saf/amf/amfnd/su.cc
> > b/osaf/services/saf/amf/amfnd/su.cc
> > --- a/osaf/services/saf/amf/amfnd/su.cc
> > +++ b/osaf/services/saf/amf/amfnd/su.cc
> > @@ -481,6 +481,7 @@ uint32_t avnd_evt_su_admin_op_req(AVND_C
> > AVND_COMP *comp;
> >
> > /* SU has been repaired. Reset states and update AMF director
> > accordingly. */
> > + LOG_NO("Repair request for '%s'", su->name.value);
> >
> > for (comp =
> > m_AVND_COMP_FROM_SU_DLL_NODE_GET(m_NCS_DBLIST_FIND_FIRST(&su->comp_list));
> > comp;
> > diff --git a/osaf/services/saf/amf/amfnd/util.cc
> > b/osaf/services/saf/amf/amfnd/util.cc
> > --- a/osaf/services/saf/amf/amfnd/util.cc
> > +++ b/osaf/services/saf/amf/amfnd/util.cc
> > @@ -236,9 +236,8 @@ void avnd_failed_state_file_create(void)
> > */
> > void avnd_failed_state_file_delete(void)
> > {
> > - if (unlink(failed_state_file_name) == -1)
> > - LOG_ER("cannot unlink failed state file %s: %s",
> > - failed_state_file_name, strerror(errno));
> > + // file might not exist in some cases, ignore errors
> > + (void) unlink(failed_state_file_name);
> > }
> >
> > /**
>
>
> ------------------------------------------------------------------------------
> CenturyLink Cloud: The Leader in Enterprise Cloud Services.
> Learn Why More Businesses Are Choosing CenturyLink Cloud For
> Critical Workloads, Development Environments & Everything In Between.
> Get a Quote or Start a Free Trial Today.
> http://pubads.g.doubleclick.net/gampad/clk?id=119420431&iu=/4140/ostg.clktrk
> _______________________________________________
> Opensaf-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/opensaf-devel
------------------------------------------------------------------------------
CenturyLink Cloud: The Leader in Enterprise Cloud Services.
Learn Why More Businesses Are Choosing CenturyLink Cloud For
Critical Workloads, Development Environments & Everything In Between.
Get a Quote or Start a Free Trial Today.
http://pubads.g.doubleclick.net/gampad/clk?id=119420431&iu=/4140/ostg.clktrk
_______________________________________________
Opensaf-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-devel
